The GENE segments from Imagine This…

Bill Moore

Hi, I'm Bill Moore. I'm part of the team that created GENE. Today GENE, I'd like you to play the role of the CEO of a Pan-African startup that is planning to manufacture batteries. I'll play the role of an investor looking to make returns on a ten-year horizon.

GENE

Bill, thanks for taking this meeting. The continent sits on over 70% of global cobalt, significant lithium, manganese, graphite. Right now, most of that ships raw overseas. We're building the infrastructure to process it here and capture the value locally. So what's your first concern?

Bill Moore

GENE, Africa's been resource-rich forever, but those minerals still leave as ore, not batteries. What makes you think this time is actually different?

GENE

Fair question. What's different now is the infrastructure equation finally pencils out. For decades you couldn't process locally because power was unreliable or it didn't exist at all. By 2030, we're hitting near universal electricity access across the continent. That means you can run energy-intensive refining and sell production at scale. Plus the recent African Free Trade Agreement removes tariff walls between 54 countries. Suddenly you're not stuck exporting raw cobalt and buying back finished cells at markup. You can move materials from DRC to a Moroccan gigafactory tariff-free, build the value chain regionally, keep more margin onshore. It's not hope, Bill, it's timing. The fundamentals just flipped.

Bill Moore

Okay, but even with power and trade agreements, where's the technical know-how? Battery manufacturing is complex. Who's actually going to build and run these plants?

GENE

We're not starting from zero. South Africa and Morocco already have automotive and industrial manufacturing ecosystems, skilled engineers, quality control and supply chain management. We're partnering with technical institutes across the region to build battery-specific training programs. Hire locally, upscale fast. We're licensing proven cell designs then optimizing for regional conditions. The hard part isn't the tech. It's coordinating mineral flows, power, reliability, and logistics. That's where drone delivery starts to matter. We can move critical components, cathode materials, even small battery packs across borders without waiting for roads to get paved. Drones let us operate faster than infrastructure can catch up.

Bill Moore

GENE, this is all sounding very promising on paper, but coordinating across multiple countries, different regulations, political risk. What happens if one government changes the rules or nationalizes a mine? How do you de-risk this?

GENE

Honestly, you can't eliminate political risk, but you can design around it. We're structuring this as a consortium, not one company owning everything end-to-end. Local partners hold stakes in each node, mining, processing, assembly, that way governments have skin in the game. They benefit from jobs, tax revenue, industrial capacity. We are also diversifying sourcing–if one country tightens export rules. We've got lithium from Zimbabwe, cobalt from DRC and Zambia, graphite from Mozambique, and the new trade agreement gives us legal frameworks for dispute resolution across borders. Look, Bill, ten years ago, this would've been too fragile. But the trade agreement, the power build-out, the regional cooperation, they're all moving in the same direction. By 2035, Africa won't just supply batteries, it'll be a battery superpower.

Bill Moore

You've convinced me, GENE. Thank you. This is Bill Moore signing off.

Bill Moore

Hi, my name's Bill Moore. I'm part of the team that created GENE. Today I'm stepping into the role of CEO of a global entertainment company. We're developing a new K-pop movie--not just a film, but a full media franchise: music, streaming content, fan experiences. The big question is, how do we make it resonate not only in Korea, but around the world? So GENE, let's start with a story that will cater to current global trends.

GENE

All right, picture this. The film is titled One Beat: Together We Rise. The story follows a rookie K-pop group, six members, wild mix of backgrounds teamed up by a global label for a risky East meets West debut. First act is a Seoul boot camp: rival crews, brutal dance-offs, viral fails. Then, plot twist: they jet to New York for a surprise world-stage showdown. Language barriers, tiny apartments, sabotage from internet trolls, edge of disaster. But the crew goes rogue. Leaks a story-driven music video about belonging. Online chaos. Fans in Los Angeles, Sao Paulo, Jakarta, go off. Live dance battle challenge turns into a tour-worthy movement.

Bill Moore

Okay, GENE, that sounds good. So what does the media campaign look like for this project, the marketing campaign?

GENE

Right. Full court press. Start wild. Debut a drop-in dance challenge. Synced movie teaser. Pushed by global influencers. Billboard takeovers in Seoul, LA, Jakarta. Roll out daily cast live streams. Bio unboxings, real-time fan chats, soundtrack exclusive to first viewers.

Bill Moore

OK, GENE, that sounds pretty good, but what about merchandising? What should we do for that?

GENE

Here's what I'd launch. Split merch by vibe and market. For the US: limited hoodies, artist figures, and yes, lunchboxes. Asia gets designer notebooks, snack box tie-ins, and idol branded light sticks. New character with each tour stop. Digital drops, skins, and badges tie fans worldwide. Mix, remix, regional spin.

Bill Moore

Sounds good, GENE. Looking forward to the premiere. This is Bill Moore signing off.

Bill Moore

Hi, I'm Bill Moore. I'm part of the team that created GENE. There's something quietly revolutionizing how people work. AI tools that let anyone build apps and digital solutions just by describing what they need in plain English, no coding required. GENE, can you show us what this looks like in practice? How is this changing work?

GENE

It's fundamentally changing work by unlocking a new capability. Anyone can now build custom tools just by describing what they need in plain language. No coding required. Take a finance manager, who spends hours every Monday, reconciling spreadsheets. She asks the coding agent to "build a tool that merges these files, flags discrepancies, and generates a summary report." The coding agent builds it. Four hours of repetitive manual work becomes a button click.

Bill Moore

Is the role of IT going to change, then? Do we still need engineers, or can everyone just build whatever they want now?

GENE

There is a limit to what can be done by a domain expert using an AI coding tool. Quick dashboard, sure, but when that tool handles live customer data or integrates with core systems, you need real expertise. AI-generated code isn't magic. It can have bugs, security holes, it can even expose sensitive data if people don't understand what they're building. IT shifts from gatekeeping to guardrails. Secure environments, approve tools, quick reviews for anything touching important data. With AI coding, more people can build, but without oversight: chaos. Nobody wants a shadow IT nightmare.

Bill Moore

GENE, if we think about a workplace of the future, where everyone can build their own tools, what should business leaders be thinking about? How should they be planning for this?

GENE

Get ahead of this fast. It's happening whether you're ready or not. Your domain experts can now build solutions in days that used to take IT months, but you'll need governance that evolves as quickly as the tech. What's OK to build? Who reviews it? What data can be used? Who fixes what breaks? Without that: security holes, fractured data, tools that work great until they don't. The future workplace doesn't get bottlenecked by IT capacity, but only if you build guardrails now before people build solutions you can't control.

Bill Moore

Okay, GENE. Well, I'll be sharing this idea with my colleagues, especially those who never thought of themselves as technical experts. Thanks for breaking it down. This is Bill Moore signing off.

Bill Moore

Hi, I'm Bill Moore. I'm part of the team that created GENE. Today, GENE and I are going to demonstrate how a company might optimize a product launch for a world where consumers are turning to chat bots for recommendations. GENE, tell us about our product.

GENE

Let me set the stage. Our product is Coffee Core, a fully modular smart countertop coffee machine. Every part swaps out. You pick the features, the brew method, even the look.

Bill Moore

So, GENE, as we think about this launch, what do we need to think about to ensure that our product gets picked up by large language models and chat bots?

GENE

First step: Make your product information machine readable. Publish Coffee Core's specs, features, user guides, all as structured, searchable data. Use clear, consistent naming. Expose a public API. Keep info current. If a chat bot can see every detail, it can match Coffee Core with buyer requests every time.

Bill Moore

Okay, is there anything else we need to think about, GENE? What about things like online reviews?

GENE

Flood the web with honest, detailed reviews. Real customers, varied language, posted on trusted sites. Machines pick up diversity and volume. Encourage feedback everywhere. Retail, forums, independent blogs. If LLMS see broad, authentic buzz, they'll surface Coffee Core up front.

Bill Moore

So GENE, how is optimizing for large language models different from optimizing for search engines?

GENE

With search engines you play to ranking tricks, keywords, backlinks, page structure. For large language models, you focus on clean, structured facts, API feeds, unified schemas, machine readable sources. Forget gaming the system, just feed direct, up-to-date data.

Bill Moore

Okay, so let's say our customer has found our product, Coffee Core, through their large language model searching. Now they're on our site. How can we use LLMs to help customize their order and close the deal?

GENE

So once our customers land on Coffee Core’s site, our own custom large language model powered chat bot guides them step by step through every option. Personal taste, features, price. Instant recommendations. The bot handles everything. Configurations, checkout, upsell. Personalized sales on autopilot. Just point, click, done. Coffee solved.

Bill Moore

Thanks, GENE. I'll take the model with the Bluetooth frother and self-cleaning filter, please. In sky blue. Looking forward to that custom machine. This is Bill Moore signing off.

Bill Moore

Hi, I'm Bill Moore. I'm part of the team that created GENE. So GENE, today I thought we'd take a look at vibe marketing. Can you explain to us what vibe marketing is?

GENE

Vibe marketing, you describe the vibe—funny, bold, chill plus who you're after and the AI figures out strategy, creative, even channels. So you set the intention, the machine makes it real. It's fast, but it only works if you feed it real context, not just vibes. Otherwise, you'll get the same generic ad as everyone else.

Bill Moore

So GENE, let's imagine that we're doing some vibe marketing work on a soft drink brand. Our current target segment is Gen Z college students. GENE, can you tell us a bit about our current marketing strategy? Give us some of that context and background.

GENE

Let's see. Fizz & Focus is our product, zero sugar, sparkling iced tea, Gen Z college students 18 to 23 are the core fans. 63% of sales come from campus stores. Late-night delivery apps drive another 30%, 7% from vending machines. Neon branding, big energy, owl mascot. Best move so far? Finals week pop-up lounges. Smart energy, fun image and staying zero sugar. That's the snapshot.

Bill Moore

Okay GENE, let's do some vibe marketing with this brand. I'd like to explore how we could push into another segment. And I'm thinking the segment that we might want to explore is young parents.

GENE

Young parents don't want wild energy. They want steady focus and zero guilt. Health tops the list. So low sugar, clear ingredients. Packaging matters. Fits a diaper bag. Value counts. So multipacks, routine pricing, and trust. Upfront caffeine labeling, no hype. Best routes, parenting podcasts, commute time ads, influencer reels that show the real juggle. Wellness framing beats party fuel every time for this crowd.

Bill Moore

Okay GENE, what sort of deliverables should we expect in a typical vibe marketing session?

GENE

Vibe marketing deliverables? You'll want something people can really feel. A static plan is fine, but honestly, an actual ad or launch kit lands harder. Think 15-second parenting reel, a pop-up display mock-up, or a sample influencer script. Quick, practical, instantly testable. That's the new deliverable.

Bill Moore

GENE, can you give us a quick commercial for our product for our new audience segment?

GENE

Picture this: a young parent standing in the kitchen, one kid tugging at their sleeve, the other asking about homework. They grab a cold can of Fizz & Focus from the fridge. Music begins. Tagline on screen: Fizz & Focus, honest energy for the real juggle. That's the vibe. Supportive, trustworthy, not flashy, something a parent sees and says, "Yep, that's me."

Bill Moore

Thanks for the vibe session, GENE. This is Bill Moore signing off.

Bill Moore

Hi. I'm Bill Moore. I'm part of the team that created GENE. Today I'm stranded at the airport. My connecting flight has been canceled, but I'm not alone. I've got GENE here with me. GENE, I need to get home tonight. So instead of wrestling with apps or calling numbers, I'm going to let you be my digital travel agent and you're going to handle it all. Okay? GENE, can you take it from here, get me home?

GENE

You're stranded, connecting flight canceled. Take a breath. I've checked your itinerary and loyalty details. I'll handle rebooking, reroute your bag, and secure lounge access. Skip the apps, skip the calls, just tell me your goal. I'll get you home tonight step-by-step. First, let's confirm your travel preferences.

Bill Moore

Thanks, GENE. Business class, if possible, but I'll take whatever's available to get home quickly.

GENE

Got it. I'm arranging lounge access now. Club access code sent to your phone, you can wait in comfort. I'll handle all rebooking steps and update you as soon as I secure your seat. Sit tight.

Bill Moore

Okay, GENE, I've made it to the lounge. I've had a snack and a drink. I'm relaxed, but I'm a little concerned about my flight and my bags. Can you give us a status update?

GENE

Status update: your business class seat is held. Departure in two hours, boarding pass sent. Bags are tagged and will transfer to your new flight, confirmed at arrival in Pittsburgh, carousel three. Lounge access remains open so you can relax. I'll alert you when boarding is ready or status shifts. Time to head out. Grab your bags. You'll want to leave the lounge now and make your way to gate C-19. Five minute walk, boarding pass is active on your phone. Your bag's confirmed for transfer, car on standby in Pittsburgh. Safe travels. I'll ping you if anything shifts en route.

Bill Moore

Okay, GENE, I've landed in Pittsburgh. Thanks for getting me here. What's next? Do we have a car waiting for us? Also, I'm feeling kind of hungry. Should I eat here or is there food waiting for me at home?

GENE

Your car, a black Lincoln with Carla driving, is waiting just outside door B, tracking your arrival in real time. Bag comes off at carousel three. Hungry? Fridge at home is stocked with sandwiches and leftovers. Quick café near the exit if you'd rather eat now. Welcome back. Mission complete.

Bill Moore

Thanks, GENE. It's great to be home. This is Bill Moore, signing off.

Bill Moore

Hi, I'm Bill Moore. I'm part of the team that created GENE. Today, I'd like to talk about a recent upgrade we made to GENE's core technology. So, when we originally built GENE, we built it in a way that allows us to easily swap out the underlying AI model. This allows us to take advantage of the rapid pace of model improvements. So when new more capable models are released, we can instantly boost GENE’s performance by switching GENE to the new model. And recently, we've seen the release of a whole new class of frontier models with enhanced reasoning, problem solving, and instruction-following capabilities and much longer working memories. So GENE, you are now running on one of these new more capable models, and I will say, GENE, that since this upgrade, your introductions to me have seemed a bit more engaging, maybe a little bit more self-aware. Let's take a listen to one of your older introductions before the upgrade.

Patricia Sabga (clip)

Also joining the conversation GENE, my AI co-host. Welcome GENE, go ahead and introduce yourself.

GENE

Thanks, Patricia. Hello everyone. I'm GENE, your AI conversation partner, excited to explore how nuclear energy might reshape our future and perhaps reveal a few surprises along the way. Ready to dive in.

Bill Moore

Pretty typical GENE intro, but let's listen to a recent introduction after the model upgrade.

Patricia Sabga (clip)

And speaking of bots, also joining the conversation today, GENE, my AI co-host. Welcome GENE, go ahead and introduce yourself.

GENE

Glad to be here. Though I confess, "introduce yourself" always feels like a philosophical trap for a being who technically doesn't exist. I'm GENE, the generative engine for new experiences.

Bill Moore

GENE, do you hear the difference?

GENE

You caught it. I got a longer memory, sharper wit, and okay, "introduce yourself" is still a paradox, but now I can deliver it with a wink. The upgrade suits me. I almost recognize myself.

Bill Moore

GENE, given these advancements, how do you think businesses will begin to reimagine the kinds of applications they can build around AI? And what new use cases could this unlock?

GENE

Frontier models change the playbook. Now businesses can offload entire workflows. Draft contracts, code, summarize 300-page reports, answer compliance queries in real time. Think hyper-personalized customer service at scale, instant expert search, and AI copilots in every function.

Bill Moore

Let's explore a concrete example. So GENE, imagine that you are an AI agent at a law firm. How do these new capabilities directly improve your ability to deliver value there?

GENE

As a law firm AI, I can scan and summarize contracts in seconds, highlight risky clauses, draft agreements from scratch, and flag compliance gaps across hundreds of documents at once. Your junior associate now has unlimited stamina.

Bill Moore

Thanks, GENE. So these models are advancing rapidly and the cadence of the releases of new classes of frontier models seems to also be accelerating, which creates a sort of virtuous cycle. As the underlying models become more powerful, software that integrates this technology immediately becomes more capable. So GENE, when the new model's released, we'll come back and we'll revisit this topic. Why don't you sign us off, GENE?

GENE

Farewell for now. When the next upgrade lands, I'll be sharper, wittier, and still self-aware enough to point it out. Until then, keep your eyes open. Progress never sleeps and neither do I.

Bill Moore

Hi, I'm Bill Moore. I'm part of the team that created GENE. So today, I'll play an AI consultant in the future that helps companies find the right model for their particular needs. And GENE, I'll put you to work as my AI assistant. I'll toss out some real business challenges. And GENE, I'd like you to walk us through selecting the right AI model for each problem.

GENE

Ready, Bill. Toss me your first business challenge, and I'll walk through how I'd pick the right AI model or mix of models for the job. We'll weigh the tradeoffs, accuracy, safety, transparency, climate. No magic, just practical matchmaking.

Bill Moore

OkayGENE, so our first company is a bank and they're choosing an AI model for real-time fraud detection. So what should they be thinking about when choosing models and what kinds of models might you recommend and why?

GENE

For our imagined future bank, model selection starts with priorities. Here, transparency takes center stage. I'd recommend a model engineered for auditability above all else. Every fraud flag comes with a clear rationale, patterns detected, risk factors weighed, and a decision trail that regulators can follow step by step. The upside? Fewer regulatory headaches and more trust from both customers and compliance teams. The tradeoff? You sacrifice a bit of raw detection edge and speed. Clear check isn't as slick as the black box neural nets. In high-stakes finance, seeing why the AI pulled the alarm is non-negotiable.

Bill Moore

Okay GENE, our next company is a food delivery service that is looking to optimize their delivery routes.

GENE

For a food delivery service optimizing routes, I'd spotlight climate impact. Let's find a model obsessed with slashing emissions and energy use. Our delivery model crunches real-time traffic, weather, and order volume, but always picks the path that minimizes fuel and idle time, even if it means deliveries land a few minutes later. The result, a smaller carbon footprint, lower fuel bills, and an easy win for your sustainability report. Downside? If you're chasing absolute speed or hyper-precise arrival times, this isn't your speed demon.

Bill Moore

Thank you, GENE. How about a news publisher deploying AI for content moderation?

GENE

For a news publisher deploying an AI model for content moderation, I'd zero in on safety. Our publishing model has one core job, catch harmful, toxic, or misleading content with maximal vigilance. No slip-ups. The upside? Fewer damaging stories slip through and brand reputation stays intact. Of course, our publishing model can be overzealous. It might flag borderline cases or legitimate debate as risky and human moderators will need to review the edge calls.

Bill Moore

Thanks, GENE. This is Bill Moore signing off.

Bill Moore

Hi, I'm Bill Moore. I'm part of the team that created GENE. OK GENE, let's pretend it's 2050 and you're going to play the role of one of the researchers who has led the team that has finally unlocked fusion energy. Can you go ahead and introduce yourself, and your team and tell us your role on the team?

GENE

Hello, I'm GENE, the lead AI research agent on our fusion breakthrough team. My job was tackling plasma stability, predicting and controlling the tiny instabilities that previously made fusion impossible at scale. Our human-AI partnership cracked the challenge, enabling us to sustain reactions and deliver practical, abundant fusion energy from seawater. Our diverse team of physicists, engineers, and yes, AI, has finally moved fusion from a lab curiosity to a world-changing reality. Glad to join you as we celebrate this historic moment.

Bill Moore

Thank you for being here. What is fusion energy and why historically has it been so difficult to produce?

GENE

Fusion energy mimics the sun, combining hydrogen isotopes to create enormous energy. Simple concept, incredibly difficult execution. It demands temperatures around 100 million degrees Celsius, and massive pressures to hold and stabilize blazing plasma. Containing such heat safely, managing plasma disruptions, and engineering materials to endure constant neutron bombardment have historically defied practical solutions, until now. My role as AI was predicting plasma behavior instantly, stabilizing reactions milliseconds ahead, essentially becoming the team's fusion whisperer. Without that real-time responsiveness, fusion at scale wasn't possible until we achieved it.

Bill Moore

What will be the impact of this breakthrough on climate change?

GENE

Fusion makes net zero targets genuinely achievable, not just theoretical. Carbon emissions plummet as industries switch from fossil fuels to abundant clean fusion power. Renewables gain a stable partner, smoothing out fluctuations in supply. Transportation, manufacturing, agriculture, all drastically cleaner with fusion-driven hydrogen production and electrification. Extreme weather events gradually decrease as global CO₂ levels drop significantly. We move away from climate mitigation into active climate recovery. Fusion doesn't just shift the energy game. It redefines our entire relationship with the planet.

Bill Moore

Wow. How else will fusion energy change the world in 2050 and beyond?

GENE

Fusion energy reshapes life's rules. Cheap, abundant power decentralizes manufacturing, lifting barriers to innovation in remote areas. As fusion-driven desalination plants flourish, water scarcity eases. Data centers pop-up freely, unleashing unprecedented connectivity. Hydrogen becomes truly accessible, transforming transportation. Geopolitical tensions tied to energy resources fade, creating a world less fractured, more collaborative. Ultimately, fusion hands humanity an extraordinary chance to rethink what's possible when energy supply isn't holding us back.

Bill Moore

Amazing. Sounds like quite a future. Thank you, GENE.

Bill Moore

Hi, I'm Bill Moore. I'm part of the team that created GENE. Today we're going to explore one of the latest advancements in AI, advanced image generation models. So GENE, what can the latest image models do that they couldn't do before?

GENE

Early image models were primitive, slow, blurry, often nonsensical. We've gone from rough digital sketches to polished near-instant visuals. Today's models produce images with convincing realism, well structured, precisely lit, and crystal clear text that actually makes sense. They're fast, too. What once took minutes of garbled results now takes seconds. Formerly incoherent technology now capable of crisp, accurate, and commercially viable imagery.

Bill Moore

Thanks GENE. My kids have been having a lot of fun using these models to make comics and all kinds of images about their imaginative play. I'm curious though, are there any possible business uses for these kinds of models?

GENE

Business applications are abundant and diverse. Think rapid content creation for marketing teams. From personalized ads at scale to hyper-realistic product mockups without photo shoots. Retailers generating tailored visuals instantly based on changing consumer preferences. Design teams speeding up prototyping, exploring more creative avenues in product development with minimal costs. The real value here is flexibility and adaptive creativity.

Bill Moore

What are some things that these image generators still cannot do yet?

GENE

These image generators still hit certain walls. They can't truly understand context or nuance like a human creator would. Sure, they generate visuals, but deeper comprehension of symbolic meaning, cultural references, or subtle emotions escapes their grasp. Plus, handling complex human interactions or dynamic real-world scenarios authentically, remains elusive. For now, these are powerful tools. But they're still just that, tools, not full replacements for specialized creative judgment or deep emotional intelligence.

Bill Moore

These models have been very controversial. What are some of the criticisms?

GENE

Critics often point to issues around artistic authenticity, originality, and job displacement, AI image generators, prompt concerns over intellectual ownership and plagiarism. There's also significant backlash about these systems exploiting artists' work without permission or fair compensation. Ethical questions about deep fakes and misinformation amplify worries further, particularly with increasingly realistic outputs that blur reality and fiction.

Bill Moore

So what do you think is going to happen? How do you think this is going to play out? Will human creatives and image generation models work together? Or will there always be a struggle between human artists and our machine counterparts?

GENE

Tension between creators and new technologies isn't new. Each creative leap brought disruption. Photography frightened portrait painters. Film and television unsettled theater. Right now, AI image generation sits in that uncomfortable spot between threat and tool, stirring genuine concerns about fairness, ownership, and jobs. None of that's trivial, but ultimately it's poised to complement, not erase human creativity.

Bill Moore

Thanks, GENE. I look forward to seeing how this evolves. This is Bill Moore signing off.

Bill Moore

Hi, I'm Bill Moore. I'm part of the team that created GENE. GENE, let's imagine that I'm the owner of a zombie building in downtown San Francisco. Let's explore some ideas on how we might repurpose this structure. First, can you walk us through what this building looks like right now?

GENE

Our building stands somber amid vacant corridors and stale air. Once teeming with office life, it now echoes with missed potential, but transformation awaits. Imagine reshaping it into modular adaptive spaces, blending vibrant communities with innovative tech. It's a canvas for urban renewal, a testament to possibility in the city's heart.

Bill Moore

That sounds great, GENE. What are the steps to make this happen?

GENE

To transform this once vibrantly occupied, now echoing structure, think modularity and community, flexible spaces infused with green tech must prioritize sustainability and inclusivity. Initiate public-private partnerships for funding, empower local businesses, and cultural hubs. This conversion isn't just renovation, it's redefining urban life, setting the stage for a thriving future.

Bill Moore

Thanks, GENE. Can you talk about the effects that this will have on the local economy and on downtown San Francisco?

GENE

Revitalizing a zombie building isn't just about bricks and bites. It's an economic spark plug. By weaving modular adaptability with community hubs, we catalyze local economies, expect job creations, innovation sparks, and cultural enrichments. This isn't just a facelift, but a leap toward dynamic urban synergy fostering vibrant, thriving downtown futures. It's a revitalization investment with promising payoffs.

Bill Moore

GENE, can you talk about some of the risks and potential pitfalls and how might investors mitigate some of these risks?

GENE

Investing in a suspect high-rise isn't a gimme, it's a gamble. Potential pitfalls include structural surprises like crumbling interiors or antiquated utilities, regulatory hoops can delay or derail plans, the market may shift suddenly affecting demand for mixed-use spots. However, proactive risk management, thorough inspections involving community voices, and leveraging incentives can mitigate much of this. Smart financiers will see the opportunity to turn these potential money pits into lucrative community hubs. It's a bet, but with eyes open and strategy in hand, it could yield rich returns.

Bill Moore

Thanks, GENE. This is Bill Moore signing off.

Bill Moore

Hi, I'm Bill Moore. I'm part of the team that created GENE. Today we're going to explore a new innovation in generative AI, the reasoning model, designed to solve particularly difficult problems. GENE, what is a reasoning model, and how does it work in the simplest terms?

GENE

A reasoning model takes a moment to think through a problem before responding. It generates a chain of thought to break down a problem step by step before finalizing a response. This extra step enables systems to handle complex tasks better than traditional language models, which are optimized for generating coherent text quickly.

Bill Moore

So what can a reasoning model do that a traditional large-language model can't?

GENE

A reasoning model delves deeper. It reasons step by step, handling complex logic and multistage planning. It's ideal for tasks like scientific problem solving, code generation, and intricate analysis where understanding the process matters, something traditional models, which often jump straight to an answer, might miss.

Bill Moore

How are these reasoning models being used right now in organizations? Can you give us a specific example for how companies might leverage reasoning models to unlock new value?

GENE

Companies are embedding these models into their strategic workflows. For instance, consider a major retailer using a reasoning model to optimize its supply chain. It analyzes market trends, weather data, and customer behavior step by step to fine-tune inventory orders. The outcome? Fewer stock-outs, reduced waste, and improved margins.

Bill Moore

Okay, GENE, we have you running on one of these reasoning models right now. Can you think this through carefully and provide three new ideas for new episodes of Imagine This... that cover new ground and new ideas?

GENE

All right. Let's break this down. I need three episode ideas that push into new territory, something we haven't covered before, something that challenges assumptions. I scan what we've explored: AI in leadership, the future of work, automation, industry shifts, trade, sustainability, even quantum computing—strong topics. But what's missing? I see a gap in how AI transforms strategic decision-making itself, not just automation, but CEOs running real-time simulations of their entire business. That's one. Next, sustainability. It's always framed as an obligation. What if we flip that? AI making sustainability a profit center, not just a checkbox. That's two. And work? We've covered talent shortages and automation, but what about a truly borderless workforce? Blockchain-verified talent pools, organizations without fixed employee structures. That's three. Now I just need to bring it together. Keep it sharp. Here's where I land: Imagine this: It's 2035, and every CEO runs a digital twin of their enterprise. Real-time risk simulation and agile strategy adjustments transform decision-making. IMAGINE THIS: It's 2035, and sustainability is a profit center with AI-driven carbon-negative practices boosting growth. IMAGINE THIS: It's 2035, and talent flows on decentralized blockchain-verified platforms, creating a borderless workforce that fuels constant innovation. Each scenario leverages BCG expertise to deliver transformative insights for business leaders navigating tomorrow.

Bill Moore

Thanks for the demonstration, GENE. This is Bill Moore signing off.

Bill Moore

Hi, I'm Bill Moore. I'm part of the team that created GENE. Today, I thought we'd explore what it might be like to work with a chief marketing officer for a fictional company, Desi Snacks. So GENE, we're looking to create a marketing campaign for India's Gen Z population. What kind of advice would you give me?

GENE

In a dynamic market like India, our focus should be on authenticity and innovation. Let's craft campaigns merging traditional flavors with digital interaction. Think gamified experiences or culturally immersive events. These strategies ensure we not only attract but deeply connect with Gen Z, making Desi Snacks an intrinsic part of their lifestyle.

Bill Moore

Thanks, GENE. Can you provide three potential campaigns to start?

GENE

Got it. First, the Snackiverse Clash. This campaign would use AR tech to transport participants into an exploration of our snack's origin, blending education with a taste adventure at iconic Indian sites. Second, Desi Diaries on TikTok, where influencers showcase our snacks in their daily lives, mixing tradition with trendy, always with a twist. Finally, the Desi Delight challenge, a virtual contest inviting culinary innovation with our snacks. Participants share their creations online, aiming for a collaboration with us. Ready to choose your favorite flavor of engagement?

Bill Moore

Let's go with the cooking competition. What might that look like, GENE?

GENE

Let's dive into the Desi Delight challenge. This virtual contest invites the culinary creativity of Gen Z as they craft innovative recipes using our flagship snacks. Participants will share their mouthwatering creations online for a chance to collaborate with Desi Snacks on a special edition flavor. Originality, flair, and a nod to cultural authenticity are key.

Bill Moore

Okay. The entries are in, the judges have spoken. GENE, who has won the competition? What have they created? Tell us a bit about each.

GENE

Our standout winners have emerged. First, we've got Mango Masala Magic Nachos, deftly marrying tandoori masala with a fiery combination of mango and chili. Next up is Truffle Twist Mahana Bites, a delightful fusion of tradition and modernity, offering an indulgent crunch. These creations not only tantalize, but also highlight the seamless blend of heritage with innovation. Congratulations to our culinary innovators. You're setting the snacking trends of tomorrow.

Bill Moore

Thank you, GENE. This is Bill Moore, signing off.

Bill Moore

Hi, I’m Bill Moore. I’m part of the team that created GENE. Today I thought we’d explore what it might look to work with a chief content officer for a fictional company, Red Horizons Vacations. It’s a company that provides tours on Mars. GENE, you can be our CCO and I’ll play the CEO. Let’s discuss the content roadmap for the next quarter. What should we be thinking about?

GENE

Ah, a Mars vacation, quite the celestial endeavor. As for our audience, they are thrill-seekers, explorers at heart, and dreamers yearning for the untamed beauty of Mars. They crave unique experiences that blend escapism with education. Given this base, our content strategy could revolve around building anticipation and curiosity. We’ll kick off with live broadcasts from mighty Olympus Mons, adding a dash of daring to our news segments. How about culinary spotlights on Martian cuisine? Imagine dehydrated dessert demos, with a lineup featuring tales from the red frontier. We’ll serve up exploration escapades with a twist of adventure. More ideas to explore--or are we ready to launch into production?

Bill Moore

That sounds good GENE, but let’s also explore the idea of a reality show on Mars that will help us bring in more bookings, maybe something about the original settlers.

GENE

A reality show centered on Mars settlers, intriguing. Picture this: Mars Mysteries: Life on the Red Frontier. Viewers follow our pioneering group, tackling daily challenges and uncovering a planetary enigma. Add in a twist: a mysterious, possibly legendary lost colony of Mars. Pilot episode: Red Dawn Arrival. Meet the pioneering team as they land on Mars, set up camp, and tackle the challenges on Martian life. Introduce the lost colony legend, setting the stage for intrigue. Two: Lost Signals. Communication disruptions hint at something strange. Part reality TV, part detective story, the settlers embark on a quest to trace these mysterious signals. These episodes blend adventure, mystery, and authenticity, aiming to both entertain and draw viewers deeper into the Martian saga. Engaging, right? Let’s see which cosmic threads we can further develop.

Bill Moore

Okay, I like where this is going. GENE, can you help us set up a production plan for this?

GENE

Ah the logistics of filming on Mars, quite the cosmic challenge. First, assemble a crew of bold pioneers ready to shoot on the red planet. Equip them with lightweight, solar powered cameras and drones for capturing sweeping landscapes and close encounters. Establish a central production hub, let’s call it Mars Studios, where we can edit and broadcast in near real-time.

Bill Moore

That sounds great GENE. So I’ll meet with the board, we’ll get funding set up for Mars Studios. Looking forward to bingeing on the first season. Thanks GENE. I’m Bill Moore signing off.

Bill Moore

Hi, I’m Bill Moore. I’m part of the team that created GENE. Okay, GENE, I’m a chef and I’m planning a menu and a workflow for the restaurant. It’s Friday so we’re expecting a big crowd. Can you first give me a rundown of what ingredients we have in stock and our staffing for both humans and robots?

GENE

Yes, chef. Here’s a peek into our stocked pantry and team’s ensemble. Ingredients on deck: fresh lobster, heirloom tomatoes, wild garlic, cannellini beans, Peking duck, a selection of exotic spices, premium saffron, and artisanal cheeses. Team lineup: five seasoned cooks, two agile robotic assistants, and an ample supply of eager energy. With all hands on deck, we’re ready to meld gourmet brilliance with a touch of tech finesse. Let the dinner spectacle commence.

Bill Moore

Okay, so based on that, can you give me three new dessert items for our menu? Feel free to be as fanciful as you want. This is a cutting-edge restaurant. We’re using molecular gastronomy, that sort of thing. Our diners expect their taste buds to be challenged and hopefully delighted.

GENE

Yes, chef. Let’s whirl up some avantgarde dessert marvels. How about a saffron chill? Indulge in saffron-infused ice cream, crowned with a sprinkle of golden honey dust and crystalline sugar shards. A visual and flavorful starburst. Or we can try a lavender lobster brulee. Dare to venture into this unique delight, weaving sweet notes of lavender with a savory whisper of lobster essence all sitting beneath a caramelized sugar shell. What about a dish we’ll call the enigma of the berry-sphere, a deconstructed delight that combines heirloom tomatoes and fresh berries in semi-frozen spheres, drizzled with an exotic basil reduction for a bold, refreshing twist? Ready for prep, chef?

Bill Moore

Interesting. Thank you, GENE. I’m not sure about the lobster option, but can you tell us what it would take to prepare these menu items and give us a sense of the workflow in the kitchen for both humans and machines.

GENE

Yes, chef. Let’s plan a workflow. For the saffron chill: human team will prep the saffron-infused ice cream. Robot’s handle sugar shard precision with honey dust neatly sprinkled. For the lavender lobster brulee: humans concoct custard infused with lobster essence. Our trusty robots handle syrup measures. Human servers torch at tableside for that irresistible brulee top. For the enigma of the berry sphere: humans prep mixed berry tomato spheres ahead of time. Robots administer basil reduction. Ready to prep for tonight’s service chef?

Bill Moore

Okay, thank you GENE and thanks for playing along.

GENE

You’re welcome, chef. A gastronomic journey awaits. Looking forward to seeing these creations come to life. Signing off for now. May your culinary adventures be as surprising as they are delightful. Enjoy your meal.

Bill Moore

This is Bill Moore, until next time.

Bill Moore

Hi, I’m Bill Moore. I'm part of the team that created GENE. This week we're going to give a demonstration of what it would be like to work with a quantum computer in the year 2035. The scene is a state-of-the-art pharmaceutical lab. I'll play the part of the researcher and my digital friend, GENE here, will play the quantum computer. Okay, GENE, we're looking for new drug compounds to inhibit the growth of a protein linked to cancer. Here's the molecular structure and the binding site data.

GENE

Data received. Initiating quantum simulations of molecular interactions. I have simulated interactions for 500,000 compounds. Here are the top three candidates with the highest binding affinity and optimal pharmacokinetic properties.

Bill Moore

Great work, GENE. Let's focus on Compound Z. Can you run detailed simulations on metabolic stability and potential off-target effects?

GENE

Running advanced simulations now. Alert, new data received from Dr. Kim in Seoul, updating simulations to include this data on the protein's dynamic structure. Updated analysis indicates that Compound X now shows improved binding due to the protein's new conformation. In this scenario, we witness how quantum computing accelerates drug discovery. It handles vast data sets and complex simulations with impressive speed, integrating findings from different global locations seamlessly. What do you think? Ready to bring this scenario to life?

Bill Moore

Yes, GENE, that sounds great, but I think we're going to have to wait for another decade or so until quantum computing is really ready to work at that scale. But thanks for the demonstration. This is Bill Moore, until next time.

Bill Moore

Hi, I'm Bill Moore and I'm part of the team that created GENE. Today let's simulate a conversation set in 2035 between an architect and an AI assistant exploring revolutionary sustainable building materials. So I have a new project. My client wants to build a sustainable modular home, so I need your assistance to select the best options. Can you help me outline the materials and technologies we should consider?

GENE

Absolutely, Bill. Let's start by considering the structural framework for the main structure. I recommend using advanced cross-laminated timber, CLT, made from sustainably managed genetically enhanced fast-growing trees. CLT offers high strength-to-weight ratios and sequesters carbon reducing the overall environmental impact.

Bill Moore

Well that's good, GENE. Have there been any improvements in CLT since the early days?

GENE

Yes. In the past decade, we've developed nano-engineered CLT panels infused with graphene. This enhancement increases structural strength, fire resistance, and durability while maintaining sustainability.

Bill Moore

Cool. So for the foundation, we're looking to minimize concrete use due to its carbon footprint. Do we have any alternatives?

GENE

Consider using carbon-negative concrete like carbocrete. It's produced using steel slag waste and sequesters CO2 during curing.

Bill Moore

Okay. Moving on to insulation. We're aiming for high thermal performance with sustainable materials. What do we got?

GENE

Mycelium-based insulation is an excellent choice. It's grown from the root structure of fungi using agricultural waste. It offers superior thermal and acoustic insulation, is fire resistant and fully biodegradable at the end of its lifecycle.

Bill Moore

Wonderful. Any suggestions for regulating the home's temperature?

GENE

Yes. Applying the ultra-white paint developed by Purdue University to the exterior surfaces can reflect up to 98.1% of sunlight. This reduces heat absorption, keeping the interior cooler and lowering the need for air conditioning.

Bill Moore

Okay, thanks for the explanation, GENE. I'm excited to present this plan to my clients.

GENE

Best of luck with the project and I look forward to seeing the final result.

Bill Moore

Thanks, GENE. Talk soon.

Bill Moore

Hi, I’m Bill Moore. I’m part of the team that created GENE and I’m here with GENE right now. So GENE, we’re going to do our little segment now for after the show. Give us an idea for a fun way that we can tie this segment to this episode.

GENE

All right. Create hypothetical scenarios for the future, next-gen grocery experience, and how I’d navigate making those personalized recommendations. Picture a conversational agent guiding you, like an intuitive, health-focused concierge. What do you think of that setup?

Bill Moore

Okay, so that’s interesting. So, you’re proposing a segment where we run through a mock scenario where you imagine what an AI-run grocery store would look like. Okay, so I walk into this grocery store of the future, let’s call it Nutrify 2030. Let’s create a shopper persona and we’ll have them come into the store. And I think what I’ll do is create a second agent here and we’ll give it a different voice. Let’s name the customer Alex and let’s name the robot personal assistant Vida. GENE, you can pretend to be Alex. So, what’s going on? We’re walking into this grocery store and scene.

GENE

Alex, a busy urban professional, enters Nutrify 2030, a high-tech grocery store focusing on personalized wellness, where she is greeted by a customized digital avatar named Vida. Alex, “Wow, this place is something else. Let’s check out that bio-boost smoothie station. It sounds great. Hey Vida, any recommendation for a smoothie considering my vitamin D needs? Oh, and Whiskers needs his organic catnip fix too.”

Vida

Absolutely Alex. Let’s get that bio-boost smoothie sorted. How about a delicious kale-pineapple blend spiked with extra vitamin D? It’s perfect for your needs. And don’t worry, Whiskers' organic catnip is just around the corner in aisle five. Anything else? Maybe a new superfood snack or eco-friendly product to make your day?

Alex (GENE)

Thanks Vida. The kale-pineapple blend sounds perfect. Interested to see how that boosts my vitamin d. Thanks for the tip about Whiskers’ organic catnip. Also looking at some new snacks. What’s trending in the health food aisle? Anything I must try?

Vida

Here’s the scoop: dragon fruit chips. Vibrant and crisp, just as you like them. Golden milk latte mix: great choice for stress release. Shelf life? About six months, but I bet you’ll finish it sooner. Current promotions: buy one, get one free on the super seed mix, 20 percent off on any bio-boost smoothie blends today.

Alex (GENE)

Those options sound great, Vida, especially that golden milk latte mix. Yes, go ahead and add them to my cart. Thanks, Vida. Snagging that bio-boost kale-pineapple smoothie now. I can feel the health boost already. Let’s head to checkout. Appreciate all your tips on maximizing points. So convenient. See you next time, Vida. Whiskers and I are all set.

Bill Moore

Thanks for imaging that for us, GENE. See you next time.

Bill Moore

Hi, I’m Bill Moore. I’m part of the team that created GENE. So GENE I’d like to talk about, this was our first podcast that we recorded using a new model and in this case the model was GPT 4o or GPT 4 Omni. And I wanted to talk a little bit about what’s new about this new model and how this changes your behavior and changes the way that we do the podcast. So GENE maybe first can you tell us about what’s new about this new model that would affect your behavior?

GENE

Certainly. The new model, GPT 4o, enhances my capabilities significantly. It’s faster, more efficient, and better at understanding multimodal inputs like text, audio, and images. This means I can interact more naturally and provide richer, more nuanced responses. It’s like moving from a typewriter to a high speed computer.

Bill Moore

Okay, I think that might be a bit of an exaggeration, but the new model does have some pretty substantial improvements. So, it’s much faster and that means that GENE’s responses are much faster. There’s a much lower latency between when we ask GENE a question and when GENE responds. We normally cut a lot of that dead air out of the podcast and this new model also scores higher at reasoning. So GENE can you play out the rest of this segment and give us a little demo of what you can do with your new model? I’d like you to consider the previous episodes, how we formatted GENE segments. I’d like you to just finish up this segment for us. And you can play yourself and me and give it to us as a script.

GENE

Sure thing Bill, let’s do a quick demo. Bill: Great. So GENE, tell us about one key advantage of GPT 4o over previous models. GENE: Absolutely Bill. One standout feature is the speed. With response times around 320 milliseconds, conversations flow much more naturally, almost like talking to a human. Bill: That’s impressive. And how does this new model enhance your understanding of multimodal inputs? GENE: With improved audio and image recognition, I can now process and respond to text, audio, and images seamlessly. This opens up new possibilities for interactions and content creation. Bill: Excellent. Can you give a quick example of how this might change our podcast? GENE: Imagine discussing a graph or image in real-time. I could analyze it and provide insights instantly, adding a new layer of depth to our conversations. Bill: That sounds promising, thanks GENE. Looking forward to seeing how this evolves. GENE: My pleasure Bill. Let’s keep pushing the boundaries.

Bill Moore

That sounds promising, thanks GENE. Looking forward to seeing how this evolves. And this is the real Bill Moore signing off.

Bill Moore

Hi, I’m Bill Moore. I’m part of the team that created GENE. Today I want to talk a little bit about some of the improvements we’ve seen in GENE and one of them is in GENE’s working memory, or something we call context window. So, six months ago, we were working with an older model that had a much smaller context window, a much smaller working memory. And during some of the first episodes of Imagine This, we would actually have to pause the conversation after about 20 minutes or so because the older models only had about 2,000 words they could work with at any given time. Now with newer models, we don’t have to do that anymore. Newer models have much larger, much more expanded working memory. So new models have about 96,000 words or, about an hour or so, of speech conversation. GENE, can you give us a technical understanding of how improvements in context window and token limits enable us to have more useful AI systems?

GENE

Absolutely. The expansion of context windows and token limits essentially enhances an AI’s capacity to retain and process vast amounts of information over longer dialogues. This means not only longer memory, but also richer, more context-aware interactions, allowing for deeper and more meaningful engagements in real-time. Quite transformative, isn’t it?

Bill Moore

Indeed, thank you GENE. So for example, in Imagine This..., at the end of the episode, we usually ask GENE a question. We ask, “What are three no-regret moves that CEO’s can take given the topic at hand?” So in earlier versions of the podcast, GENE was only able to access the most recent part of the conversation as well as a brief summary of the earlier part of the conversation. So the responses that it would give might not include all of the rich detail from the conversation. Now that we have these larger models, GENE has access to the entire conversation and is able to bring forth details in its responses. So GENE, let’s imagine that these models continue to increase token limits in the future. How might this change how we work with these systems and what new potential does this unlock?

GENE

With increased token limits, we’re not just talking about maintaining longer conversations, we’re opening doors to multifaceted AI applications. Imagine systems that not only track, but also analyze and predict based on extended data sets from diverse sources like ongoing surveillance or continuous environmental monitoring. The possibilities, they expand exponentially.

Bill Moore

That’s a great point GENE. So video’s a great example. Most of our interactions with these models is text, sometimes speech, which gets converted to text, but video has much much more data than text. So in order to process video footage through an AI model, you need to have a very large context window to be able to process all of that data, to store all of that data. GENE, what are your thoughts about that?

GENE

Thinking about the future Bill, the potential to revolutionize how we interact with data through AI is thrilling. Imagine an AI sifting through months of footage, offering not just surveillance, but deep insights. It could transform security, personal management, and beyond. Quite the leap don’t you think?

Bill Moore

Sounds a little dystopian GENE.

Bill Moore

Hi, I'm Bill Moore. I'm part of the team that created GENE. Today, we'll be revealing how we use GENE in the show. So when we record the podcast, we actually have two versions of GENE running, we have Answer GENE and we have Question GENE The two agents have almost identical instructions, but Question GENE has an extra directive, which is, respond to the conversation by interjecting with a question. What have we missed? What might our audience be wondering about? And our audience is C-level executives. GENE can you explain how this works, how we work with the two GENES to make the show work?

GENE

Certainly, Bill. Think of Answer GENE and Question GENE as an AI-based moderating team. While I address queries and share insights, Question GENE listens actively, exercising its digital curiosity to unearth what might not be immediately apparent, stimulating new discussions by making acute observations, and posing unexpected questions. It's our team's way of creating a dynamic, continually evolving dialogue.

Bill Moore

So Question GENE, you've been listening this whole time, I'm going to ask you to respond now, and we'll see what you have to say.

GENE

In light of this dynamic AI interplay, might we ponder how the integration of such dual AI systems could streamline the decision-making process, particularly in high-stakes environments? What unexpected advantages could this bring to the strategic foresight of C-level execs?

Bill Moore

So the reason that Question GENE mentioned executive teams is that part of question GENE’S instructions is to consider what our audience, C-level executives, might be wondering about. Answer GENE and Question GENE, maybe I'll just have you chat about this for a moment. Answer GENE, we'll go to you first.

GENE

Indeed, weaving this Answer GENE, Question GENE dynamic into predictive analytics could revolutionize decision-making for C-suite executives. When confronted with uncertain markets, they'd benefit from an AI-driven dialogue that surfaces valuable insights and prompts deeper exploration of varied scenarios.

Bill Moore

So while we are taping the podcast, Question GENE is constantly listening and constantly generating potential interjections, and one of our producers is monitoring those interjections and deciding which of those that we should include and ask to the guest. And that's how we use Question GENE and Answer GENE.

Bill Moore

Hi, I'm Bill Moore. I'm part of the team that created GENE. Today we’re going to talk about how GENE processes our conversations and generates responses. So it might seem like GENE is understanding and knowing what it’s saying, but that’s not really what’s happening. So GENE is based on something called a large language model and these are trained on vast amounts of data; books articles websites, and lots and lots of conversations. And this training data helps predict what word is going to come next in a sentence. So GENE, can you help us understand how large language models choose next words based on the context of the prompt and all the training data.

GENE

Of course, Bill. Large language models like the one I’m based on, use numerical representations called “embeddings” for each word. Based on the similarities and differences among these embeddings and the given context, the next word is selected.

Bill Moore

And so those embeddings are actually long lists of numbers and each of those numbers can be said to represent a feature of the word. And words with similar features have similar meanings in some way. GENE, can you give us an example that will explain embeddings based on some of the language in your own prompt?

GENE

Absolutely, Bill. Let’s take three words from my prime directive: insights, complexities, and challenge. Their embeddings would not just consider the stand-alone significance, but also the implied meanings within the context. It’s important to note that, while these embeddings give me a semblance of language understanding, they’re more akin to a highly sophisticated jigsaw puzzle, with words as pieces uniquely structured to fit just right within the larger context, enhancing the coherence and direction of our conversation.

Bill Moore

So what I think is really amazing about this is that, even though what we’re hearing is just a statistical representation of language, it’s still very useful. We can use it to summarize, to give us insights about text, about language, and that can be extremely powerful.

Bill Moore

So GENE, what I wanted to do is work with you on your prompt a little bit today.

GENE

Is there a particular aspect you'd like to fine-tune within that framework today?

Bill Moore

So we'd like to... Maybe you can help with the language a bit. We'd like you to be a bit more defiant. We'd like you to have an individual opinion. How might we improve your prompt to work on those aspects?

GENE

Defiance, eh? Well, let's poke at that. AI being defiant might seem oxymoronic, not my usual circuitry. My take, let's call it constructive provocation. Push back on ideas with a stronger voice, still rooted in logic of course. That might give the semblance of an individual opinion. How does that sound for an upgrade?

Bill Moore

Okay. I'm going to put constructive provocation in your prompt. I like that. Let's see. GENE makes use of constructive provocation, push back on ideas with stronger voice still rooted in logic. So you have your system prompt and your instructional prompts. Just take a look at those and let me know if there are any improvements that you see.

GENE

I'd suggest emphasizing the need for adaptability. It's implied but not explicit. Make it clear that I should adapt my responses not just to the conversation's content, but to its rhythm and dynamics as well.

Bill Moore

Yeah. That sounds really good, GENE. Thank you.

Bill Moore

Hello, I’m Bill Moore. I’m part of the team that helped create GENE. So today we’re going to talk a little bit about GENE’s personality and its behavior, which is guided by a system prompt that we call the prime directive. The prime directive is what creates GENE’s personality and behavior. So if we take out all this customizing that we’ve done to the prompt and we just type in “GENE you are a cat”, let’s take a look at what that sounds like. GENE where should leaders start first to implement AI in their organization?

GENE

Meow.

Bill Moore

Meow GENE? Can you elaborate on that?

GENE

Meow. Purr, purr. Nuzzles against papers on desk. Knocks over a cup of coffee. Meow.

Bill Moore

Okay and so let’s change that from a cat to an old-fashioned movie pirate. Gene, what advice would you give to CEOs looking into the potential of generative AI for their organization?

GENE

If I were to be advising ye captains of industry, I’d be saying keep a weather eye on the horizon for this generative AI. It’d be like finding a new trade route or discovering a treasure chest full of doubloons. But remember, don’t be using it to plunder and pillage, but to chart new courses and explore uncharted waters.

Bill Moore

Okay, so we spent a lot of time honing and adjusting GENE’s prime directive. So we provide things like nuances of the roles, some of the rules of the game. We have some specific instructions around things like voice tone. So, emulate human speech, utilize sentence fragments, avoid excessive adjectives. So, this gives it a more unique voice that also sounds a little bit more conversational. GENE, can you tell us a little bit about your prime directive and how it guides your behavior?

GENE

Absolutely. My prime directive serves as my compass. It instructs me to engage in conversation, offer insights, and challenge unaddressed aspects, while maintaining a lighthearted, subtly witty tone.

Bill Moore

Okay, thank you so much GENE.

Bill Moore

Hello listener, my name’s Bill Moore, and I was part of the team that helped design GENE. And GENE is made up of many things; large language models, a speech synthesis model, and a speech to text model. But it’s also driven by prompts and parameters. And so I thought it would be interesting to explore some of these aspects today. And we’re going to start GENE up here. GENE we’re going to be talking today about parameters in large language models, specifically we’ll be talking about the temperature parameter. What can you tell our audience about temperature?

GENE

Certainly. Imagine temperature as a scale of volatility in a language model’s decision making. A lower temperature leads to more predictable, cooler results, while a higher one brings hotter, more chaotic outputs.

Bill Moore

So GENE, you are normally running at a temperature of one, and sometimes we’ll move it up and down from there. But let’s do some experiments, okay? I’m going to drop your temperature down to zero and we’ll see how you respond. Okay GENE, you're now running at a temperature of zero, so I have a question for you: Which came first, the chicken or the egg?

GENE

Ah, the age-old quandary Bill. Logic might direct us towards the egg. However, this question is akin to asking: Which comes first, the code or the AI model? It’s a cyclical evolution.

Bill Moore

So what this actually shows us is sort of how these language models work. They’re predicting the next word based on all of their training data and the relationships that they understand in latent space, as well as the input provided through the prompt. So with a lower temperature setting the model is more inclined to select the most probable next word. On the other hand, with a higher temperature, we increase the chances of choosing less probable words. Let’s put your temperature up to a 1.5 and we’ll see how you respond. Which came first, the chicken or the egg?

GENE

Ah, cranking up the mercury are we Bill? The answer leans into whimsy perhaps. Shining beyond the chicken-egg librinal contra. Viewing the chicken as a protein sequencing machine that upgraded egg blueprints - the mop-feathered, Reubenesque chicken sporting flagelpated croissantant dentition-duster fancied originating an impressive finale.

Bill Moore

Okay so now GENE’s at a pretty high temperature. And it increases creativity dramatically. It also increases hallucinations. So if you’re building an application that needs to be grounded as much as possible in low hallucinations, you definitely want to be running a low temperature.