What the AI?! – Details, episodes & analysis

In this episode of What the AI?!, Jeff and Annie dive into the explosive debate around China’s DeepSeek AI model, which wiped out a trillion dollars in U.S. market value in a single day. They also explore a Harvard study showing that AI tutors can more than double student engagement and learning gains, and the latest update from the U.S. Copyright Office, which ruled that AI-assisted works can be copyrighted (with a few conditions)

In this Episode:

• Meta’s new AI personalization feature and what it means for users’ privacy.
• The potential risks of AI-generated misinformation and its growing impact on global politics.
• How AI is changing enterprise solutions, from automating workflows to improving customer interactions.

Relevant Links:

• Unitree H1 robots perform Yangge dance at China's New Year Gala.
• AI-powered dancing robots showcase China's advances in robotics.
• Harvard AI tutor doubles student engagement in physics.
• DeepSeek R1 model shakes up AI efficiency.
• Nvidia Project Digits and open-source innovation.
• AI tutor significantly boosts learning gains.
• AI surpasses traditional teaching in engagement.
• Tesla Roadster launched into space in 2018.
• Tesla mistaken for an asteroid by astronomers.
• Meta AI’s personalization enhances user experience.
• OpenAI launches ChatGPT Gov for U.S. agencies.
• OpenAI collaborates with U.S. National Labs.
• AI-assisted works require human creativity for copyright.
• Copyright Office clarifies AI-generated content rules.

This week on What the AI?!, Jeff and Annie unpack the details behind Trump’s headline-grabbing $500 billion investment in U.S. AI infrastructure, featuring tech titans like Oracle, OpenAI, and Softbank, raising tough criticism. They also discuss how open-source AI models, like DeepSeek’s R1, challenge industry giants in the AI space, and explore the privacy vs. productivity battle between Apple and Android as AI agents take center stage.

We also discuss:

• The Oscars controversy around AI-enhanced accents and its implications for artistic recognition.
• The remarkable success of AI tutoring in Nigeria, where students advanced two years in just six weeks.
• The use of AI to design anti-venoms for snakebites, a revolutionary step in medical research.

Relevant Links:

• Respeecher’s AI voice-editing enhanced Hungarian accents in "The Brutalist."
• AI scores snowboarding tricks at the 2025 X Games.
• DeepSeek-R1 outperforms OpenAI’s o1 with a cheaper, open-source model.
• OpenAI, SoftBank, Oracle: $500B U.S. AI data center investment.
• Apple disables AI news summaries after backlash from false headlines.
• Google Gemini updates enable advanced multimodal tasks, revolutionizing AI assistants.
• Perplexity Assistant launches Android-first multimodal, third-party app integrations. 
• OpenAI “Operator” browser agent automates tasks like online shopping.
• Nigerian students achieve two years of learning in six weeks with AI tutoring. 
• AI-designed antidotes neutralize snake venom, revolutionizing medicine.

 Is AGI closer than we think? Or are bold predictions just hype?

Artificial Intelligence is making headlines again—from robots trained with virtual simulations to AI-assisted music earning Grammy nominations. But with predictions of Artificial General Intelligence (AGI) by 2025, are we truly on the verge of this breakthrough, or are we letting speculation overshadow reality - and does it really matter?

In this episode of What the AI?!, we explore the latest AI news, debate the feasibility of AGI, and share practical ways to harness today’s AI tools in your daily life.Key Discussion Points:

• AI Meets Music: The Beatles’ “Now and Then” becomes the first AI-assisted song to earn Grammy nods.
• Google’s New AI Tools: Exploring AI for video creation, and the new Gemini app for iPhone
• Robot Training Breakthroughs: How MIT’s LucidSim is revolutionizing task accuracy through virtual simulations.
• Scientific Research with AI Bots: How Stanford researchers got multiple AI chatbots to effectively collaborate on scientific challenges
• The AGI Debate: Are 2025 predictions realistic, or is true AGI still decades away?
• Hands On: Jeff shows how LLMs can interpret and visualize numbers, turning anyone into a data analyst.

Join us for a grounded look at the latest in AI and what it means for the future.

In this episode of What the AI?!, Jeff Keltner and Annie Delgado take you on a journey through the latest twists in AI policy, tech breakthroughs, and what they mean for the future. Curious about AI’s role in elections? They dive into Perplexity’s Election Information Hub, where Jeff shares his firsthand experience and explores the risks and benefits of AI-driven voter info.

They’ll also cover some game-changing tech: a real-time Minecraft world generator that adapts to player movement, and OpenAI’s $15 million acquisition of the chat.com domain, stirring up questions on whether branding could be a game-changer for AI dominance.

Plus, they’re breaking down Google’s Gemini API, which now integrates real-time Google search to cut down on AI “hallucinations” and serve up grounded, accurate info—could this be the future of search?

And for a bigger picture, Jeff and Annie tackle the regulatory maze surrounding AI. As policymakers weigh in on data privacy, algorithmic bias, and potential over-regulation, Jeff and Annie discuss the tension between keeping AI fair and not crushing its potential. With real-world examples and a look at EU vs. U.S. approaches, this episode is packed with insights on the fine line between innovation and protection.

Whether you’re an AI enthusiast or just curious about the headlines, this episode has something for everyone.

Welcome to "What the AI"—your front-row seat to the latest in AI breakthroughs for business leaders and executives on the go.

In this episode, Jeff Keltner and Annie Delgado break down the week’s biggest AI developments and the game-changing tools redefining productivity. From Perplexity’s new knowledge search that integrates private databases to OpenAI’s factuality benchmark, they dive into how LLMs are moving from basic assistants to real agents that act on your behalf.

ChatGPT’s web search for paid users? Covered. Anthropic’s Cloud that can actually control your computer? Yep, we’re talking about it all. Plus, they explore the hype around Salesforce’s rebranded “AgentForce” and why the term “agent” might mean much more than just a fancy chatbot.

Are we ready to trust AI to plan vacations, manage finances, and even handle the fine print on legal documents? Jeff and Annie unpack these questions, blending excitement with caution as they imagine a future where AI does more than just chat—it takes action.

In this episode of What the AI?!, Jeff and Annie explore groundbreaking advancements in AI technology. From ChatGPT’s new “Tasks” feature, which aims to revolutionize personal and professional productivity, to Columbia University’s GET model unlocking gene expression mysteries. Plus, Nvidia’s bold ventures in healthcare include partnerships with Mayo Clinic and cutting-edge AI-driven pathology tools. It’s a packed episode highlighting the transformative power of AI in science, medicine, and everyday life.

In this episode:

• How ChatGPT’s Tasks feature brings AI agents closer to reality—but still has room to grow.
• Columbia’s GET model and its potential to unlock the mysteries of human gene expression.
• Nvidia’s sweeping partnerships in healthcare and what they mean for the future of medicine.
• The growing tension between AI innovation and global policies—what new restrictions could mean.
• Why AI’s potential to solve complex real-world problems is more exciting than ever.

Every week, we see AI advances growing at an exponential pace.

Are these breakthroughs paving the way to solve humanity’s greatest challenges—or creating our biggest threats?

In this episode of What the AI?!, Jeff and Annie dive into Nvidia's groundbreaking world models and their rivalry with DeepMind, exploring the next leap toward AI that bridges the longtime issue of physics and reality in AI. They uncover the chilling potential of AI-powered spear phishing scams, the groundbreaking promise of brain-computer interfaces, and how AI is saving lives by revolutionizing cancer detection and diagnosis.

In this Episode:

• The ethical and societal questions raised by brain-computer interfaces.
• How AI-powered simulations could revolutionize robotics and video gaming.
• The economic implications of open-source models vs. proprietary AI tools.
• What Microsoft’s $80 Billion AI bet means for America’s tech future

Relevant Links:

• Nvidia Cosmos World Models Announcement  
• Microsoft Phi-4 Model on Hugging Face 
• AI Spearfishing Research 
• Microsoft’s $80 Billion AI Investment Announcement 
• Meta’s AI-Powered Profiles Update 
• NeuroXess Brain-Computer Interface Development 
• AI in Cancer Screening Study 

What’s next for AI in 2025?

In this episode, we kick off the year exploring how ChatGPT spiced up Christmas traditions, the IRS’s use of AI to combat fraud, and what personalized AI tutors could mean for education in the future. Plus, we dive into Meta’s ambitious plans for AI-generated characters and the most exciting gadgets (and flops) from 2024.

We also discuss:

1. ChatGPT in holiday fun and New Year’s resolutions.
2. How AI is revolutionizing fraud detection, with the IRS leading the charge.
3. The bold future of education with AI tutors and hands-on workshops.

Relevant Links:

• IRS AI Tools to Fight Fraud
• DeepSeek-V3 Model Announcement
• Stanford AI Model for Avatar Gestures
• Meta’s AI Characters Announcement
• AgiBot World Alpha Dataset
• ARMOR by Carnegie Mellon and Apple
• OpenAI Public Benefit Corporation Transformation
• OpenAI Deliberative Alignment Safety
• Arizona Unbound School Launch
• MIT Technology Review on AI Gadgets

AI has hit a wall. Can it pass through it? 

With claims of “peak data,” the era of more-is-better in artificial intelligence might be over. What’s next could reshape how AI innovation unfolds.

In this episode of What the AI?!, Jeff Keltner and Annie Delgado tackle one of the most pressing questions in AI: what happens when we reach “peak data”? As OpenAI’s Ilya Sutskever suggests, AI models can no longer rely on simply consuming more text-based information to get smarter. Instead, the future may belong to reasoning-driven systems that think like humans, such as Meta’s COCONUT model and Google’s Gemini Flash. They also explore cutting-edge tools like Veo 2.0, which pushes the boundaries of video realism, and Whisk, Google’s playful new image generation tool.

In this Episode:

• Google’s Veo 2.0 video generation model surpassing OpenAI’s Sora with realistic object permanence and physics.
• OpenAI’s accessibility push with tools like 1-800-CHATGPT and free internet search for all users.
• Whisk, Google’s new image generation tool blending creativity by combining uploaded styles and themes.

Relevant Links:

• Google Labs VO 2.0 Tool
• OpenAI’s 1-800-CHATGPT Tool
• Google Labs
• Ilya Sutskever’s NeurIPS Keynote on Peak Data