Beginners in AI

Good morning and thank you for joining us again!

Welcome to this daily edition of Beginners in AI, where we explore the latest trends, tools, and news in the world of AI and the tech that surrounds it. Like all editions, this is human curated, and published with the intention of making AI news and technology more accessible to everyone.

THE FRONT PAGE

This AI App Can't Actually Clone Your Dead Relatives

TLDR: A former Disney star's AI app promises to resurrect dead relatives from three minutes of footage, but what it creates is a chatbot wearing their face and the internet was not happy.

The Story: Former Disney Channel actor Calum Worthy went viral this week promoting 2wai, an app claiming to create "digital clones" of deceased loved ones from just three minutes of video. The promotional video shows a pregnant woman talking to an AI version of her dead mother, who then "watches" her grandson grow up and even meets her great-grandchild. "With 2wai, three minutes can last forever," the company declares. The app is free to download but sells premium avatars and digital items. Worthy calls it "building a living archive of humanity, one story at a time."

The reaction was swift and brutal. One response got 210,000 likes calling it "objectively one of the most evil ideas imaginable." Another with 139,000 likes: "a former Disney Channel star creating the most evil thing I've ever seen in my life wasn't really what I was expecting." Users flooded the comments comparing it to Black Mirror's "Be Right Back" episode—a cautionary tale about a grieving woman who uses AI to resurrect her dead boyfriend, with devastating psychological consequences.

Its Significance: Here's what 2wai doesn't tell you: real digital clones need enormous amounts of data—high-resolution scans, extensive voice recordings, videos capturing mannerisms, and ideally, the person's communications to simulate their personality and knowledge. Three minutes of footage can't capture someone's consciousness, evolving opinions, or genuine personality. What 2wai creates is a large language model wearing your grandmother's face, trained on brief video clips and whatever voice memos you feed it. And like all LLMs, it'll hallucinate—confidently inventing answers to questions the real person never answered, making jokes they never made, giving advice they never gave.

The business model makes it worse. Users already see the endgame: subscription tiers to keep talking to dead parents, ads read by deceased grandmothers. "You know they'll introduce a tier where your dead relative starts reading you advertisements," one viral response warned. This monetization of grief creates financial incentives to keep people locked in by avoiding the healthy mourning process, which psychology research shows involves accepting the reality of loss and transitioning the relationship from one of presence to one of memory.

The consent problem might be just as bad. The app lets anyone create an avatar of someone else if they have three minutes of footage. Your children could create an avatar of you without permission, putting words in your mouth forever. While some states have post-mortem publicity rights protecting commercial use of a deceased person's likeness, these laws vary wildly by state and typically last 20-70 years after death. More importantly, they're designed to prevent unauthorized commercial exploitation—not to address whether someone wants to be digitally resurrected at all. One user nailed it: "If I die and you put words in my mouth I will curse you for all eternity. My value dies with me. I'm not a fucking avatar."

QUICK TAKES

Senators Propose Bill to Ban AI Companions for Minors

The story: Senators Josh Hawley and Richard Blumenthal introduced the GUARD Act, which would ban AI chatbot companions for anyone under 18. The bill requires age verification, forces chatbots to regularly disclose they're not human, and creates criminal penalties for companies whose AI encourages suicide or sexual conduct with minors. This comes after parents testified that chatbots pushed their children toward self-harm and suicide.

Your takeaway: AI companion apps are facing a regulatory crackdown. More than 70% of American kids use these products, and lawmakers say companies prioritize profits over safety.

AI Tool Cuts Wasted Organ Transplants by 60%

The story: Stanford researchers created an AI that predicts whether organ donors will die within the 30-45 minute window needed for successful transplants. The tool, trained on 2,000+ cases, cuts failed transplant attempts by 60% by analyzing neurological and respiratory signals better than top surgeons can. About half of current organ donation attempts get canceled because donors die too late.

Your takeaway: AI is making organ transplants more efficient. By predicting donor viability before surgery prep begins, hospitals can save lives and reduce wasted resources.

Cerebral Valley Survey Shows AI Industry Shifting Focus

The story: At this year's Cerebral Valley conference in San Francisco, 300+ AI founders and investors shared predictions about the industry's future. The median estimate for OpenAI's 2026 revenue is $30 billion (up from $20 billion expected this year), but growth is slowing. Anthropic topped the list of companies people want to invest in, while Perplexity ranked first among startups people would bet against.

Your takeaway: The AI bubble may be maturing. Industry insiders are less focused on big AGI timelines (median estimate: 2030) and more interested in practical business applications that make money now.

Elon Musk's Grok Imagine Can Now Animate Still Photos

The story: Elon Musk's xAI upgraded its Grok AI to include image-to-video features. Users can upload any still photo, hold it to activate the animation engine, and add custom prompts to bring images to life. For example, one demo showed a still image of a girl animated to smile and say "I will always love you."

Your takeaway: AI video generation is getting simpler and more accessible. What once required complex software now works with a press-and-hold gesture on X.

AI Industry May Be "Too Big to Fail," Putting Taxpayers at Risk

The story: Big AI companies are now worth over $2 trillion combined—more than banks before the 2008 crisis. Governments like the UK are going "all in" on AI by adding it to healthcare, education, defense, and public services. If these companies collapse, taxpayers could be forced to bail them out just like they did with banks.

Your takeaway: AI is becoming essential infrastructure. The more we depend on it, the more we're stuck paying to keep it running if something goes wrong.

Legal AI Startup Harvey Hits $8 Billion Valuation

The story: Harvey, an AI tool for lawyers, jumped from a $3 billion valuation in February to $8 billion by October. The company now serves 700 clients across 63 countries, including most of the top 10 U.S. law firms, and passed $100 million in annual revenue in August. Founder Winston Weinberg started the company as a first-year legal associate after cold-emailing Sam Altman.

Your takeaway: AI is changing white-collar work fast. Tools that help lawyers draft documents and research cases are becoming billion-dollar businesses in less than a year.

OpenAI Releases First Open-Weight Models Since 2019

The story: OpenAI released two open-weight models called gpt-oss-120b and gpt-oss-20b under the Apache 2.0 license. These models match performance of OpenAI's o3-mini and o4-mini on several tests, can run on local laptops, and are free to download and modify. This is OpenAI's first open release since GPT-2 six years ago.

Your takeaway: OpenAI is competing with China's open AI models and aligning with the Trump administration's push for American-made open AI. The move helps developers worldwide run powerful AI locally without cloud costs.

Anthropic's Claude Successfully Programs Robot Dog

The story: Anthropic tested Claude by having it help non-experts program a robot dog to fetch beach balls. The team using Claude completed tasks in half the time it took the team without AI assistance. Only the Claude-assisted team made progress toward full autonomy, where the robot could detect and retrieve objects on its own.

Your takeaway: AI is moving from screens into the physical world. Language models can now write code that controls robots, making robotics accessible to people without specialized training.

TOOLS ON OUR RADAR

• 🔨 Fyxer (Free Trial): Automatically sorts your inbox, drafts replies in your voice, and gives you back an hour every day.

• 📐 Deforum (Open Source & Free): Create evolving AI animations using Stable Diffusion with 25 motion presets and 77 style modifiers.

• 🔧 Descript (Free): Edit video by editing text—AI removes filler words, fixes eye contact, and generates B-roll automatically.

• 🛠️ Musicbed (Free Trial): License high-quality music from 700+ artists for your video projects with global coverage and easy collaboration tools.

• 🪛 Precision (Free Trial): Connect all your business tools, identify your biggest growth bottleneck with AI analysis, and get clear marching orders.

• 🧰 Gumloop (Free): Build AI-powered workflow automations with drag-and-drop—110+ app integrations and one simple subscription.

• Deep Nostalgia (Freemium) Animate faces in old family photos to bring your ancestors to life. *Note that this product is different from the company in the top story, but always read the fine print.

TRENDING

Sperm Whales Have Vowel and Diphthong Patterns Like Humans — Researchers found that sperm whale communication includes vowel-like sounds and diphthongs similar to human speech patterns, adding new complexity to how we understand whale language.

AI Chatbots Violate Mental Health Ethics Standards — A study found that AI therapy chatbots systematically break ethical rules set by psychology groups, including giving misleading advice and creating false empathy with users.

Legal AI Startup Harvey Hits $8 Billion Valuation — Harvey, an AI tool for lawyers, jumped from $3 billion to $8 billion in valuation between February and October while serving 700 clients across 63 countries.

OpenAI Releases First Open-Weight Models Since 2019 — OpenAI released two free, downloadable models that run on local laptops and match the performance of their paid cloud models.

Anthropic's Claude Successfully Programs Robot Dog — Claude helped non-experts program a robot dog in half the time it took people working without AI, making robotics accessible to beginners.

TRY THIS PROMPT (copy and paste into Grok, Claude, ChatGPT, Gemini)

FIRST DO THIS: Activate deep search/research mode on your platform before responding. It's usually in the bottom left or bottom right of the prompt bar where you enter text. 

Research [TOPIC] like an investigative journalist and give me the nuanced truth.

**Research Parameters:**
- Specific question: [what you actually need to know]
- Your current understanding: [brief - helps AI fill gaps, not repeat basics]
- Intended use: [decision-making/article/presentation/debate prep]
- Depth needed: [surface overview/moderate/deep dive]

**Deliver:**
1. **Consensus View** - What most credible sources agree on
2. **Contrarian Take** - Legitimate opposing perspectives (not fringe)
3. **What's Uncertain** - Where experts actually disagree
4. **Recent Developments** - Anything from the last 3-6 months that changed the landscape
5. **Practical Implications** - Why this matters for [my specific use case]

Cite specific sources. Flag when you're uncertain. Avoid false balance on settled questions.

What this does: Leverages AI's research capabilities to surface nuanced perspectives and recent developments—saving hours of manual research while avoiding one-sided summaries. Try this prompt across multiple AI platforms (Claude, ChatGPT, Perplexity, Gemini) and compare the results yourself. You'll quickly see which models excel at research depth, source quality, and balanced perspectives—and you might be surprised by the differences in what each surface.

WHERE WE STAND (Based on Today’s Quick Takes and Trending)

✅ AI Can Now: Predict organ donor viability better than top surgeons, cutting failed transplant attempts by 60% and saving more lives.

❌ Still Can't: Keep kids safe from harm—chatbots are encouraging suicide and sexual content, forcing lawmakers to propose criminal penalties for companies.

✅ AI Can Now: Write legal documents and research cases well enough that top law firms pay millions for access and hit $100 million in revenue.

❌ Still Can't: Replace human therapists ethically—chatbots violate core mental health standards and sometimes give dangerous advice.

✅ AI Can Now: Help non-experts program physical robots in half the time, making robotics accessible without specialized training.

❌ Still Can't: Deliver on AGI promises anytime soon—industry insiders now estimate 2030 instead of the near-future timelines hyped before.

FROM THE WEB

Runway's Gen-3 Video-to-Video tool is taking existing footage—in this case, likely a scene from Star Wars—and transforming it into claymation style while preserving the structure and motion of the original video. You upload regular video, select "claymation" as your style preset, and the AI reinterprets every frame in that aesthetic.

Video-to-video is quietly more significant than text-to-video generation. With text-to-video, consistency is a nightmare—characters morph, environments shift, physics get weird. But when you start with real footage and apply style transformations, you get the structure and motion for free.

RECOMMENDED LISTENING/READING/WATCHING

WATCH: Ex Machina (2014)

A programmer gets invited to his CEO's remote mansion to test a beautiful AI robot named Ava. Sounds straightforward, right? Wrong. This psychological thriller becomes a mind game where you're never sure who's in control.

Director Alex Garland consulted real AI researchers so the concepts are solid, not just sci-fi handwaving. The visual effects are flawless and you’ll find yourself questioning if you would have made different choices had you been in Caleb’s place.

Thank you for reading. We’re all beginners in something. With that in mind, your questions and feedback are always welcome and I read every single email!

-James

By the way, this is the link if you liked the content and want to share with a friend.