Human-like, real world intelligence up to a decade away

Welcome to the first edition of The Revolution AI weekly newsletter!

AI continues to revolutionize our digital world, but the physical world presents a whole new set of problems.

Mark Zuckerberg’s Meta is working on these, but the man tasked with delivering human level AI is not underestimating the challenge.

Also, Apple offers security researchers a million dollar payday.

• Meta AI Chief: Up to 10 Years for Human-Like, 3D Intelligence

• Disney Entertains AI in Film and TV Production

• Apple Offers Hackers $1M

• From the (AI) Toolbox

• A question for you…

Human-like, “world model” intelligence within a decade

According to Meta's AI chief Yann LeCun, today's AI models can't actually think like humans, despite claims by some AI labs. In a talk at the Hudson Forum, he said current large language models like ChatGPT lack true abilities such as reasoning, and intuition.

If you’ve been watching closely, you’ll be familiar by now with the criticism that AI models are limited to predicting the next word or pixel. State of the art artificial intelligence cannot perceive or interact with the three dimensional world like humans do. As a result, even the latest AI equipped robots struggle with basic physical tasks that even young children can grasp quickly.

So while today's AI like ChatGPT seems increasingly intelligent, we aren't quite at artificial general intelligence yet. LeCun says achieving human-level competence requires building 3D "world models" - in some ways like the mental models that humans rely on.

But Meta’s founder and CEO has one burning question for his AI Chief: “Mark Zuckerberg keeps asking me how long it’s going to take”.

LeCun cautions that building these models is a challenge that will take years, and possibly even another decade to overcome.

  Entertainment

Disney Reportedly Expanding AI for Film and TV

The move signals Disney's embrace of AI under CEO Bob Iger, who has compared it to Walt Disney's use of tech for storytelling. However, previous entertainment AI projects have faced backlash. Fans and industry creatives argue AI results are noticeably inferior and threaten jobs.

As AI capabilities improve, success likely hinges on blending AI subtly into human-led creative processes. If done right, AI could help Disney tell stories more efficiently across mediums. One thing is for certain - the initiative's unveiling will be closely watched by the entertainment industry.

  Roundup  

Money Talks, AI Walks

Mega-cap tech firms like Microsoft, Alphabet, Amazon, Meta, and Apple report earnings this week. The updates could confirm the "AI use case phase" has begun in the enterprise world.

Mark Zuckerberg's pivot from the metaverse to AI is paying off, with Meta's AI investments driving sales growth and stock price gains. While the company hasn't abandoned its metaverse dreams, Zuckerberg is touting AI as the future of the internet.

Microsoft could receive substantial equity in OpenAI as the AI startup restructures into a for-profit company, according to The Wall Street Journal. Both firms have hired banks to negotiate Microsoft's stake, which follows its nearly $14 billion investment.

Generative AI startups received $3.9 billion in venture capital funding across 206 deals in Q3 2024, per PitchBook data. However, surging demand for computing power could prolong coal plant operations and increase emissions.

The Evident AI Index reveals a widening gap between AI leaders and laggards in banking, with North American giants like JPMorgan and Capital One dominating while challengers play catch-up—signaling AI proficiency is now vital for survival in modern finance.

  Security

Apple Offers $1M to Ethical Hackers

Private Cloud Compute was envisioned to enable more advanced AI while keeping data processing on users' devices, rather than Apple's servers.

With payouts of up to $250,000 also pledged for identifying flaws that could leak customer data, Apple is signaling the priority it places on safeguarding users' AI information.

By proactively hunting for problems in its private cloud, Apple is working to ensure its AI progress doesn't come at the cost of user privacy. As AI permeates consumer tech, that's a tradeoff more companies need to evaluate.

Apple's largest single bug bounty payout so far is $200,000.

From the Toolbox

• Cogito: Analyzes customer service calls, enhancing agent performance and customer satisfaction.

• Mint: AI-powered personal finance tool for budgeting, tracking expenses, and managing investments.

• Affectiva: Analyzes facial expressions for market research, improving ad effectiveness.

• H2O.ai: Automates data science tasks, accelerating model development and deployment.

• AIVA: Composes personalized music for various media, enhancing user experiences.

News

Quick Bites

• Google takes aim at supercharging web browsing with Project Jarvis, its forthcoming AI agent for automating tasks in Chrome.

• Apple's AI is pumping the breaks on "Apple Intelligence" rollout plan.

• Seeking companionship in silicon, lonely humans find solace in empathetic AI chatbots.

• Google eyes the AI agent race with "Project Jarvis" to automate web tasks.

• Meta breaks new ground with Self-Taught Evaluator, an AI that trains itself.

• Perplexity defends itself against News Corp’s lawsuit, claiming media giants fear AI's potential and prefer a world where facts are corporate property.

A question for you…

We’d also love to hear any comments or feedback. Feel free to reply, we value every email.

Thanks for reading…

See you next time!

Dan @ The Revolution AI