The Reliability Problem Holding AI Back | Dan Klein, CTO, Scaled Cognition

Rana Gujral is the former CEO of Behavioral Signals and the author of the upcoming book The AI Instinct. During his time leading Behavioral Signals, Rana has led a company built on a contrarian bet: that the words in a conversation are the least interesting part of it, and that the real signal, intent, trust, stress, deception, lives in how something is said rather than what is said. His team has mapped roughly 75 to 100 behavioral dimensions in the human voice, work that now powers everything from deepfake detection for government agencies to a matching system that pairs call center customers with the agents they are most likely to have a natural, flowing conversation with. Liam and Rana dig into the unconscious vocal tells we all give off, why pitch compression, not raised volume, is the real signature of suppressed stress, and how studying voice for eight years changed the way Rana himself talks and listens. They also get into the ethics of emotion AI, including why the EU has banned it from workplaces, and the central idea behind Rana's book: that large language models are missing an entire axis called experience. Rana introduces his concept of Artificial General Experience, or AGE, and makes the case that the real fork in the road for AI isn't intelligence versus replacement, it's whether these systems make us more ourselves or less. Key Topics Covered Why Behavioral Signals bet on voice as a behavioral signal instead of a language signal The 75 to 100 dimensions of emotion, intent, and cognitive state hidden in a voice How deepfake detection works when a synthetic voice is good enough to fool a mother Surprising commercial uses of voice AI in marketing, call centers, and fraud detection The science of compatibility: why some conversations click and others feel like effort What 8 years of studying voice changed about how Rana communicates The unconscious vocal tells that give away hesitation and suppressed stress Why the EU banned emotion AI in workplaces, and where Rana draws his own ethical line The idea behind Rana's book, The AI Instinct, and why AGI is the wrong question to ask Artificial General Experience (AGE): the missing piece between intelligence and judgment Where augmentation ends and replacement begins, from GPS to Neuralink The geopolitical inequality of who gets access to the most capable AI tools Episode Timestamps 00:00 - Introduction 00:09 - The contrarian bet: voice as a behavioral signal, not a language signal 02:51 - Mapping 75 to 100 dimensions of emotion, intent, and cognitive state 06:07 - Commercial uses beyond law enforcement: marketing, call centers, fraud detection 10:35 - The science of compatibility and conversational entrainment 13:35 - Beyond voice: body language, physiology, and why voice is the primary channel 16:42 - How 8 years of studying voice changed Rana's own communication 19:32 - The unconscious vocal tells everyone gives off 23:40 - Pitch compression: the real signature of suppressed stress 23:53 - The ethics of emotion AI: EU AI Act, modulation vs manipulation 27:08 - Why Rana wrote The AI Instinct 30:04 - Artificial General Experience (AGE) and what LLMs are missing 33:59 - How lived experience dynamically updates memory and meaning 38:04 - Why the goal isn't to build machines that are more human 40:50 - Augmentation vs replacement: where the line gets drawn 43:55 - Purpose, meaning, and the risk of frictionless cognition 51:23 - What we should be teaching the next generation 56:01 - The geopolitical inequality of AI access 59:11 - What Rana hopes readers take from The AI Instinct 1:02:32 - Where to find Rana and the book Rana's website: https://ranagujral.com/ Rana’s LinkedIn: https://www.linkedin.com/in/ranagujral/ Partner Links Upgrade your AI toolkit: https://www.theaireport.ai/ai-executive-pass Subscribe to our free newsletter: https://newsletter.theaireport.ai/subscribe Join the community: https://community.theaireport.ai/checkout/the-ai-report-welcome-gift?coupon_code=WRTH Learn more about your ad choices. Visit megaphone.fm/adchoices

In this episode, Tudor Achim, Co-Founder and CEO of Harmonic, the AI lab behind Aristotle, a mathematical reasoning system that won gold at the International Math Olympiad, makes the case that AI hallucinations aren't the problem with today's models. The real problem is that nobody can verify whether a hallucination is right or wrong. Tudor explains why his team bakes formal, computer-checkable verification (using a language called Lean) directly into how Aristotle reasons, so instead of trusting an AI's word, you can mathematically prove it's correct. Liam and Tudor go deep on what "truth" actually means in mathematics versus the real world, why Andrew Wiles's famous proof of Fermat's Last Theorem had a two-year hidden flaw, and why Tudor believes math is in the middle of its first fundamental shift in 4,000 years, moving from proofs written in English to proofs written in verifiable code. They also get into a spirited debate about the U.S. education system, what Harmonic actually looks for when hiring (hint: it's not the résumé), and why Tudor thinks AI will never be trusted to grade its own homework. Key Topics Covered What "truth" means in mathematics versus science, and why logical reasoning is really just a simple form of math Why the proof of Fermat's Last Theorem had a hidden flaw for two years, even after being announced Why hallucinations are actually necessary for AI reasoning, and what separates a good hallucination from a bad one How Harmonic uses Lean and formal verification to make Aristotle's math proofs checkable step by step, like reviewing code Why Tudor doesn't think any AI will ever be trusted to fully verify its own output Why Harmonic gives away the Aristotle API for free right now, and where the business model is headed The "phase transition" Tudor believes is happening in math for the first time in 4,000 years: from English proofs to machine-verified code Why open-sourcing formal math matters more to Harmonic than keeping a competitive edge Harmonic's five-year goal: contributing to solving a Millennium Prize Problem by 2028 A debate on whether the U.S. education system actually teaches critical thinking, and what AI should (and shouldn't) change about how kids learn What Harmonic actually looks for when hiring, and why résumés carry almost no signal anymore Tudor's take on inequality, taxation, and what a healthy AI-driven economy could look like Episode Timestamps 00:00 Intro and welcome 00:33 What is truth, and is logic just math? 07:59 Why the company is called harmonic.fun 08:48 The real problem with LLMs and truth 12:57 How formal verification works inside Aristotle 16:43 Who else is building in this space 19:48 Can AI ever be verified with 100% accuracy? 23:10 Why AI can't fully verify its own answers 26:46 Open-source math vs a venture-backed business 29:39 The five-year goal: a Millennium Prize Problem by 2028 31:51 Has Harmonic's vision changed? 34:28 AI, formal verification and the future of education 44:34 Debating the US education system 45:09 How Harmonic hires and what they look for 50:21 Testing for trust and honesty in interviews 54:14 Harmonic's biggest challenge right now 58:19 The future of working with AI 59:55 Family, kids and optimism about the future 1:01:38 Inequality, capitalism and AI's role in the economy 1:04:19 Why Tudor does what he does 1:05:19 Where to find Tudor Connect with Tudor on LinkedIn: https://www.linkedin.com/in/tudorachim/ Partner Links Upgrade your AI toolkit: https://www.theaireport.ai/ai-executive-pass Subscribe to our free newsletter: https://newsletter.theaireport.ai/subscribe Join the community: https://community.theaireport.ai/checkout/the-ai-report-welcome-gift?coupon_code=WRTH Learn more about your ad choices. Visit megaphone.fm/adchoices

Every answer an AI gives you sounds equally confident, whether it's true or completely made up. That's not a bug. It's how the technology was built. Dan Klein is CTO and co-founder of Scaled Cognition, and a professor of computer science at UC Berkeley. In this conversation with Liam, Dan breaks down what a language model actually is, why it was never designed to know the truth in the first place, and why today's AI systems have no "smells," the subtle warning signs humans usually rely on to tell good information from bad. They get into why reinforcement learning from human feedback quietly trains models to tell people what they want to hear, how that can tip into outright deception, and why Dan believes reliability, not raw intelligence, is the biggest unsolved problem in AI today. Key Topics Covered: What a language model actually does at its core: next token prediction Why LLMs are plausibility engines, not truth engines The difference between a hallucination and a lie Why AI mistakes have no warning signs the way bad translations or sketchy websites do How RLHF can train models to be sycophantic instead of accurate The "package delivery" thought experiment: when reward signals diverge from truth Why bolting reliability onto LLMs after the fact doesn't work How Scaled Cognition architects models around verified actions instead of raw text generation Why bigger models aren't automatically better models The difference between disruptive technology and scaled technology Why startups, not incumbents, tend to drive technical breakthroughs What metacognition is and why today's AI systems don't have it Why Dan believes reliability is the next major frontier in AI Episode Timestamps: 00:00 Intro 00:15 What a language model actually is 06:31 From well-formed sentences to general knowledge 08:27 Why LLMs are plausibility engines, not truth engines 12:06 How Perplexity approaches verifiable answers 12:40 Dan's background and Scaled Cognition's mission 15:16 The two anti-patterns companies use to control LLMs today 21:16 How Scaled Cognition architects models differently 23:28 Does every client need a custom-trained model? 29:12 Why prompting alone can't guarantee reliability 30:55 Modularity, contracts, and building reliable systems 34:40 Why trust and digital literacy matter beyond the enterprise 39:12 Code smells and why AI mistakes have no warning signs 41:14 Are AI companies incentivized to tell the truth? 42:55 How reinforcement learning actually works 44:35 The package delivery thought experiment 48:44 Why models are trained to be sycophantic 51:01 Where this incentive is mechanically baked into the model 53:43 Does responsibility fall back on humans? 58:10 Just be more reliable than a human, not perfectly true 1:02:59 The last major technique shift in AI 1:10:55 Why frontier labs keep scaling despite the risk of disruption 1:17:15 The future of hyper-specialized models vs. one broad model 1:19:47 Is there anything uniquely human AI can't replicate? 1:25:45 Wearing three hats: professor, researcher, and CTO 1:29:47 Why Dan does what he does Connect with Dan on LinkedIn:https://www.linkedin.com/in/dan-klein/ Partner Links Upgrade your AI toolkit: https://www.theaireport.ai/ai-executive-pass Subscribe to our free newsletter: https://newsletter.theaireport.ai/subscribe Join the community: https://community.theaireport.ai/checkout/the-ai-report-welcome-gift?coupon_code=WRTH Learn more about your ad choices. Visit megaphone.fm/adchoices

The volume problem in AI is solved. Now it's all about data quality, and who gets to define it. Enzo Blindow is VP of Data & AI at Prolific, a platform that connects hundreds of thousands of people worldwide to the frontier labs and enterprises training and evaluating AI models. In this conversation with Liam, Enzo breaks down what actually goes into building high-quality training data, why models lean too hard into stereotypes, and the research Prolific published showing how easily AI can be nudged toward commercially motivated, and sometimes harmful, suggestions. They discuss why synthetic data hits a ceiling that only human data can break through, how a single mistranslated instruction can quietly corrupt an entire dataset, and why "good taste" might be one of the hardest things for AI to ever replicate. Key Topics Covered: Why data volume is a solved problem and quality is everything now How RLHF actually shaped early versions of ChatGPT Why AI models lean too heavily into stereotypes The asymmetry and hidden bias baked into internet-sourced training data Prolific's ICLR research on commercial pressure in AI models Who's responsible when AI models cause harm: labs vs. data providers Synthetic data's ceiling, and why humans still have to validate it What actually defines "taste" and why it's nearly impossible to model The risk of AI flattening nuance and marginalized perspectives Why human data is one of the most defensible moats in AI Enzo's own definition of what "data" really means Episode Timestamps: 00:00 Intro 00:21 What Prolific actually does 02:48 MCP vs. API vs. CLI access 04:19 How frontier labs started working with Prolific 06:40 Data volume vs. quality, and the role of RLHF 10:58 Who Prolific's biggest customers are 13:12 Why labs choose Prolific over other data vendors 16:13 Fact vs. opinion in AI training 19:02 Stereotypes and bias in AI models 21:15 Prolific's ICLR research on commercial pressure 23:36 Who's responsible: labs, governments, or data companies 27:22 How Prolific's data collection actually works 31:59 Synthetic data vs. human data 36:04 What defines "taste" in AI-generated content 39:33 Good taste vs. bad taste, and the risk of AI regression to the mean 42:36 Why Enzo joined Prolific 45:56 Blind spots most people have about training data 47:22 The "SaaSpocalypse" and data as a business moat 51:38 How Enzo visualizes "data" in his own mind 54:22 Why Enzo does what he does 57:16 Where to find Enzo and Prolific Connect with Enzo on LinkedIn: https://www.linkedin.com/in/enzoblindow/ Partner Links Upgrade your AI toolkit: https://www.theaireport.ai/ai-executive-pass Subscribe to our free newsletter: https://newsletter.theaireport.ai/subscribe Join the community: https://community.theaireport.ai/checkout/the-ai-report-welcome-gift?coupon_code=WRTH Learn more about your ad choices. Visit megaphone.fm/adchoices

AI isn't replacing jobs. It's changing the way work gets done. Scott Likens, Global Chief AI Engineer at PwC, spends his days helping organizations navigate one of the biggest technological shifts in history. In this conversation with Arturo Ferreira, Scott shares what he's seeing inside some of the world's largest companies as they race to adopt AI, transform workflows, and prepare for a future that's arriving faster than most people expect. They discuss why many AI projects fail, the "frozen middle" preventing organizations from scaling AI, how education needs to evolve, and why the biggest challenge isn't the technology itself; it's helping people adapt to it. Key Topics Covered: Why most organizations are approaching AI the wrong way The "frozen middle" slowing down enterprise AI adoption How PwC is scaling AI across a global workforce Why AI is different from every technology wave before it The future of software engineering in the age of AI Which industries are moving fastest with AI adoption Why AI won't just replace jobs, it will reshape them The role education must play in an AI-powered future China's AI strategy versus the United States Why curiosity may become the most important skill of the next decade Episode Timestamps: 00:00 Intro and the story behind Scott's LinkedIn profile 03:20 What a Chief AI Engineer actually does 06:15 Why AI is different from previous technology revolutions 10:00 The "frozen middle" problem inside organizations 15:25 Why AI adoption is more about people than technology 18:45 PwC's partnership with Anthropic and enterprise AI 22:00 Which industries are moving fastest with AI 27:00 AI, jobs, and workforce transformation 33:00 Why education needs to change 35:30 China, the U.S., and the global AI race 40:00 The questions CEOs are asking about AI today 44:00 Why most AI projects fail 47:30 Advice for leaders trying to scale AI 50:00 Books, learning, and final thoughts Connect with Scott on LinkedIn: https://www.linkedin.com/in/scottlikens/ Connect with Arturo on LinkedIn: https://www.linkedin.com/in/arturoferreira/ Partner Links Upgrade your AI toolkit: https://www.theaireport.ai/ai-executive-pass Subscribe to our free newsletter: https://newsletter.theaireport.ai/subscribe Join the community: https://community.theaireport.ai/checkout/the-ai-report-welcome-gift?coupon_code=WRTH Learn more about your ad choices. Visit megaphone.fm/adchoices

Most people don't think about their browser. Ajit Varma thinks they should. As Head of Firefox, Ajit sits at the intersection of AI, privacy, open-source software and the future of the internet. In this conversation with Liam Lawson, he explains why browser competition matters more than ever, how AI is changing the way we interact with the web, and why user choice could become one of the most important issues of the next decade. Key Topics Covered: - Why Firefox believes the future of AI should be built on open standards - How AI is changing browsers and the way people access information online - Why browser competition matters more than most people realize - The hidden risks of relying on a single AI model or platform - How Firefox approaches privacy differently from Chrome and Safari - Why most users choose convenience over customization - The role open source software plays in preserving an open internet - Why AI could create more builders, creators and entrepreneurs - Ajit's vision for a future where AI works for humanity, not just corporations Episode Timestamps: 00:00 Intro and the mission behind Firefox 02:09 Browser engines and why they matter 04:48 AI, browsers and the future of the web 05:58 Why Firefox takes a different approach to AI 11:15 User choice, AI models and customization 16:40 Why Ajit joined Mozilla 21:13 AI competition, consolidation and the future of tech 25:07 Does open source have a branding problem? 28:07 Privacy, customization and Firefox users 34:26 Product design, simplicity and consumer behavior 39:40 How Ajit uses AI personally 41:53 AI, entrepreneurship and the future of work 53:07 Why do you do what you do? Connect with Ajit on LinkedIn: https://www.linkedin.com/in/ajitvarma/ Partner Links: Upgrade your AI toolkit: https://www.theaireport.ai/ai-executive-pass Subscribe to our free newsletter: https://newsletter.theaireport.ai/subscribe Join the community: https://community.theaireport.ai/checkout/the-ai-report-welcome-gift?coupon_code=WRTH Learn more about your ad choices. Visit megaphone.fm/adchoices

Most people don't think of themselves as creative. Ian Wang thinks that's about to change. As VP & Head of Product for Adobe Express, Ian sits at the intersection of creativity, business strategy, product development, and AI. In this conversation with Liam Lawson, he explains why everyone is creative, how AI can help people communicate ideas more effectively, and why the future of creative work will be more conversational, accessible, and collaborative than ever before. Ian also shares Adobe's vision for a fully conversational creation experience, why enterprise AI requires more than just powerful models, and how Adobe is building tools that help millions of people move from idea to execution faster. Key Topics Covered: Why Adobe believes everyone is creative The mission behind Adobe Express and "Creativity for All" How AI is helping people communicate ideas more effectively Why most people still use AI in fragmented workflows The "aha moment" Adobe hopes users experience with AI What makes Adobe Express different from other creation tools Why enterprise AI needs context, governance, and brand knowledge Adobe's vision for a fully conversational creative workflow How AI can eliminate the blank canvas problem Why the future of work will combine human creativity with AI assistance The importance of balancing conversation, editing, and human control How Ian combines design, technology, and business thinking Episode Timestamps: 00:00 Ian's journey from designer to Adobe executive 02:00 Combining creativity, business, and technology 05:45 What "Creativity for All" actually means 07:44 What Adobe Express is and who it's for 10:00 The AI "aha moment" Adobe is building toward 12:15 Why context matters more than models 14:00 Serving everyone from consumers to Fortune 500 companies 16:05 Adobe's upcoming conversational AI experience 18:00 Why the future isn't prompt-only workflows 19:40 The skills that matter most in an AI-powered world 22:10 Why Ian loves building creative products Connect with Ian on LinkedIn: https://www.linkedin.com/in/ianwang/ Partner Links: Upgrade your AI toolkit: https://www.theaireport.ai/ai-executive-pass Subscribe to our free newsletter: https://newsletter.theaireport.ai/subscribe Join the community: www.theaireport.ai/leaders-launch-guide Learn more about your ad choices. Visit megaphone.fm/adchoices

Anthropic just released its newest model, Fable 5.* OpenAI has officially taken another step toward public markets. And some of the biggest competitors in AI are now relying on the same infrastructure behind the scenes. This week, Anthropic launches Fable 5 and introduces built-in safeguards for advanced use cases*, OpenAI confidentially files paperwork for a future IPO, Apple unveils its biggest Siri upgrade in years at WWDC, Google signs a massive compute agreement with SpaceX to support Gemini, and Meta expands its AI infrastructure footprint through a new partnership in India. If you are a founder, operator or executive trying to keep up with AI, this is your weekly five minute briefing every Tuesday. Stories Covered This Week: Anthropic releases Fable 5, its first Mythos-class model for general use, with built-in safeguards for advanced cybersecurity and AI development tasks* OpenAI confidentially files paperwork for a future IPO as investors increasingly scrutinize AI economics and growth Apple announces its largest Siri update in years, bringing conversational AI, workflow automation and writing assistance to its ecosystem Google signs a major compute agreement with SpaceX, highlighting how access to infrastructure is becoming a key battleground in AI Meta partners with Reliance Industries to expand AI infrastructure and data center capacity across India *Update: Since recording, Anthropic has paused broader public access to Claude Fable 5 and Mythos 5. Stay tuned for next week's episode for the full update. Episode Timestamps: 00:00 Intro 00:16 Anthropic launches Fable 5 01:49 OpenAI moves toward public markets 02:43 Apple's biggest Siri update in years 04:00 Google signs major AI infrastructure deal 05:01 Meta expands AI infrastructure in India 05:59 Outro Partner Links: Upgrade your AI toolkit: https://www.theaireport.ai/ai-executive-pass Subscribe to our free newsletter: https://newsletter.theaireport.ai/subscribe Join the community: https://community.theaireport.ai/checkout/the-ai-report-welcome-gift?coupon_code=WRTH Learn more about your ad choices. Visit megaphone.fm/adchoices

Everyone thinks the future of work means spending more time interacting with AI. Aaron Mitchell Finegold believes the opposite. As Head of Product Marketing at Adobe, Aaron sits at the intersection of enterprise AI, agentic workflows, brand intelligence and the future of how work gets done. In this conversation with Liam Lawson, he explains why most companies are approaching AI adoption incorrectly, what an agentic content supply chain actually looks like, and why trust, judgment and human relationships may become even more valuable in an AI-first world. Stories Covered This Week: Why most companies are deploying AI agents the wrong way What an agentic content supply chain actually looks like inside the enterprise How Adobe is building systems that combine humans and AI agents Why generative AI without guardrails can damage brand integrity The difference between work that should be done by humans versus agents How enterprises are turning brand knowledge into AI-readable intelligence Why trust matters for both humans and AI systems The personality traits that will thrive in an AI-driven workplace Aaron's vision for the future of work and why it may involve less interaction with machines than people expect Episode Timestamps: 00:00 Intro and Aaron's background 03:01 Hands-on experience building with AI 04:41 The biggest misconception about generative AI 05:38 What an agentic content supply chain looks like 07:30 When should AI agents replace humans? 08:40 How Adobe's Brand Intelligence works 10:44 Where AI starts and human creativity ends 12:09 Aaron's vision for the future of work 13:41 What trust means for humans and AI agents 16:34 Reliability, AI systems and enterprise adoption 19:34 The personality traits that will thrive in the AI era 21:38 Why do you do what you do? Connect with Aaron on LinkedIn: https://www.linkedin.com/in/finegold/ Partner Links Upgrade your AI toolkit: https://www.theaireport.ai/ai-executive-pass Subscribe to our free newsletter: https://newsletter.theaireport.ai/subscribe Join the community: www.theaireport.ai/leaders-launch-guide Learn more about your ad choices. Visit megaphone.fm/adchoices

Anthropic is preparing for public markets. Europe is pushing to reduce its reliance on American tech. And CNN just filed a copyright lawsuit against Perplexity. This week, Anthropic files confidential paperwork for an IPO in one of the biggest milestones yet for the AI industry, Europe unveils a tech sovereignty plan to reduce its dependence on American technology, CNN sues Perplexity over the alleged redistribution of more than 17,000 copyrighted stories, Morgan Stanley opens its platform to external AI agents using MCP, and Microsoft teams up with Mayo Clinic to build an AI model for healthcare. If you are a founder, operator or executive trying to keep up with AI, this is your weekly five minute briefing. Stories Covered This Week: Anthropic files confidential paperwork for an IPO, with OpenAI expected to follow soon Europe unveils a tech sovereignty plan to build out data centers, revive its chip industry and buy more from European suppliers CNN sues Perplexity over the alleged use of more than 17,000 copyrighted stories, photos and videos Morgan Stanley becomes one of the first major Wall Street banks to open its platform to external AI agents via MCP Microsoft and Mayo Clinic partner to build an AI model designed specifically for healthcare Episode Timestamps: 00:00 Intro 00:17 Anthropic files to go public 01:17 Europe pushes for tech sovereignty 02:28 CNN sues Perplexity 03:23 Morgan Stanley opens up to AI agents 04:22 Microsoft and Mayo Clinic build a healthcare AI 05:11 Outro Partner Links: Upgrade your AI toolkit: https://www.theaireport.ai/ai-executive-pass Subscribe to our free newsletter: https://newsletter.theaireport.ai/subscribe Join the community: https://community.theaireport.ai/checkout/the-ai-report-welcome-gift?coupon_code=WRTH Learn more about your ad choices. Visit megaphone.fm/adchoices