AI News - October 2024

A quick overview of interesting AI News for the Yate and Sodbury U3A's monthly AI News meeting on October 10th 2024 (meetings are second Thursday of month, 1400-1545, St John's Church Centre, Wickwar Road, Chipping Sodbury - all welcome). Note that this is very much a summary to act as background as the topics are discussed in detail at the meeting itself. We also discuss and demonstrate other AI products and services and much more.

[1] Tesla's 'We, Robot' Day (after Isaac Asimov 'I, Robot')

Grok writes: "Scheduled for October 10, 2024, in Los Angeles, the company is set to unveil its autonomous robotaxi, tentatively named "Cybercab," marking a pivotal moment in Tesla's shift towards mobility services and advanced AI. The event will likely showcase not just the vehicle's design but also Tesla's vision for integrating robotaxis into a broader ride-sharing ecosystem, potentially including demonstrations of fully autonomous operation technology, discussions on production and operational costs, and perhaps even a sneak peek at Tesla's ride-sharing app interface. Additionally, there might be updates on other Tesla projects like the Optimus humanoid robot, hinting at Tesla's comprehensive approach to future mobility and automation. "

10/10 pic.twitter.com/cQBHqqYYtN

— Tesla (@Tesla) September 25, 2024

There are others working on self-driving of course, not least Wayve.AI which has one base in London.

London is an incredibly diverse city, full of narrow, bustling, cobbled streets and unpredictable weather - it's a challenge for any driver.
See how the Wayve AI Driver fares against these obstacles in the heart of the city. 🏙️
Full 75 Min Drive: https://t.co/KN8gUEB8A7… pic.twitter.com/gZX5YHZwl5

— Wayve (@wayve_ai) October 9, 2024

[2] Meta's powerful new video model

Not yet released, but with many powerful new features including putting personal images into videos, precise video editing through text prompts, and including sound effects and background music, which sync with the video content.

[2a] Prompts that allow amazing levels of detail

Meta's Movie Gen looks wild

Prompt: A red-faced monkey with white fur is bathing in a natural hot spring. The monkey is playing in the water with a miniature sail ship in front of it, made of wood with a white sail and a small rudder. The hot spring is surrounded by lush… pic.twitter.com/KRHbjr0Lz5

— Luis C (@lucataco93) October 4, 2024

[2b] Examples of world model and physics

The physics in the world model of @AIatMeta’s Meta Movie Gen is as good as it gets.

In fact, it’s insanely good!

It’s just unbelievable how far we’ve advanced, and how quickly! pic.twitter.com/lfnSzjqxDe

— Derya Unutmaz, MD (@DeryaTR_) October 5, 2024

[3] Google's 'World Simulator'

Google's "World Simulator" project, though shrouded in some ambiguity based on public posts and information, appears to be an ambitious venture aimed at creating a highly detailed simulation of the world or major aspects of it, possibly leveraging AI, machine learning, and extensive data sets.

[3a] Connection between video generation and world simulation

As these models scale in complexity, they develop an understanding of physical laws, object permanence, and even basic physics. This isn't because they're explicitly programmed to simulate the world but because understanding these concepts aids in producing realistic and coherent video sequences. See below how top 'text to video' people are being recruited.

I will be joining @GoogleDeepMind to work on video generation and world simulators! Can't wait to collaborate with such a talented team.

I had an amazing two years at OpenAI making Sora. Thank you to all the passionate and kind people I worked with. Excited for the next chapter!

— Tim Brooks (@_tim_brooks) October 3, 2024

[3b] Language models that are deeper than any one language

Probably points towards the existence or development of a deeper, universal logical structure beneath languages. Hence far more efficient than simple translation.

“Imagine a world where everybody has equal access to the world’s information in their own language.” - @uokon. Google has been advancing LLMs for a more language-inclusive future. Take a deeper dive into the potential of LLMs here: https://t.co/XRHShc3Tvq pic.twitter.com/wDFCfI53Cy

— Google AI (@GoogleAI) October 4, 2024

[4] Movie Studios making AI Partnerships

The top 'text to video' AI tools are partnering with the major film studios. How will they do against AI based competitors who feel they can simply make films without any traditional partner?

After signing the deal with Runway, Lionsgate said to the WSJ that they had felt pressured by impending deals being made by competitors. Sam Altman has been in discussions about Sora partnerships with the major studios since March. A studio announcement is probably imminent. pic.twitter.com/jILHaiM6CB

— Andrew Curran (@AndrewCurran_) September 21, 2024

[5] Hints that the 'Singularity' is very near

If we define the Singularity as that point at which AI has intelligence surpassing humanity, then many experts are suggesting it is coming soon. Quite what that means for humanity is rather more badly understood.

[5a] Individual statements are bold


The leaks are going to start so I might as well just come out and say it: I'm getting out of AI for good.

We're gonna be fine. There's nothing more I can do, nothing more I want to do.

Time to kick back and enjoy the ride.

I'm going to focus on writing, health, and spending…

— David Shapiro ⏩ (@DaveShapi) October 3, 2024

Also:

Deep learning worked, and AI is rapidly advancing toward superintelligence whether you're ready or not. Quit worrying about challenges; the future will be unimaginably prosperous, and there's nothing you can do to stop it.

— Flowers (@flowersslop) September 23, 2024

[5b] OpenAI on way to AGI


🚨 OpenAI researcher Noam Brown confirms that reaching AGI is our main goal

" OpenAI only primary goal is to reach AGI.

There are no single applications prioritized other than approaching the development of AGI " pic.twitter.com/dpbI80a4YU

— Haider. (@slow_developer) October 2, 2024

[5c] How the new O1 series can do advanced reasoning

The new models are able to consider their first answer and think about it before trying again and again until they reach a much better answer than their first one.


🚨 Sam Altman confirms that OpenAI has been working on O1 model for a long time

" We've been working on this model for a long time, targeting to develop AI systems that can reason through complex problems.

01 is the first to achieve advanced reasoning and it will unlock major… pic.twitter.com/53RAMYiIJi

— Haider. (@slow_developer) September 19, 2024

[5d] Sam Altman describes the 5 levels towards AGI

The Levels framework: Level 1 - Chatbots, Level 2 - Reasoners (today), Level 3 - Agents (soon), Level 4 - Innovators, Level 5 - Organisations. Agents can 'do things' like send emails, make phone calls etc. Agents will be able to be able to be 'chained together' to achieve complex outcomes.


Sam Altman at OpenAI DevDay 2024: the fact that the definition of AGI matters means we are getting close and "it's going to happen pretty quickly" pic.twitter.com/oAUmaRta0L

— Tsarathustra (@tsarnick) October 2, 2024

[6] Dangers of being left behind

Many believe that AI is advancing dangerously fast. Nevertheless, providing it does not cause great harm there are huge dangers to economies left behind. The fact that the USA has less AI legislation may mean, as below, Europe not gaining access to initial releases. The UK's position is less clear.

Advanced Voice is not yet available in the EU, the UK, Switzerland, Iceland, Norway, and Liechtenstein.

— OpenAI (@OpenAI) September 24, 2024

As a meme!

pic.twitter.com/wYzmLwkPuu

— des (@dotnetschizo) September 26, 2024

[7] Call centres about to be transformed by AI

The real time speech that was just introduced to ChatGPT can now be built into any software ('realtime API' in the post below). Clearly most call centre work can therefore now be done by AI at less cost (once initial prices fall, which will be very soon).

realtime api (speech-to-speech): https://t.co/cfUkb9hTsq

vision in the fine-tuning api: https://t.co/fDFazE7xvZ

prompt caching (50% discounts and faster processing for recently-seen input tokens):https://t.co/97mz86YVKR

model distillation (!!):https://t.co/4wH1bmP7rb

— Sam Altman (@sama) October 1, 2024

[8] AI Glasses destroy privacy

As demonstrated by the clip below, a person wearing AI glasses can video you, use face recognition software to identify you from social media, and then pretend to know you. This software is available now and the glasses cost ÂŁ300 or so.

Are we ready for a world where our data is exposed at a glance? @CaineArdayfio and I offer an answer to protect yourself here:https://t.co/LhxModhDpk pic.twitter.com/Oo35TxBNtD

— AnhPhu Nguyen (@AnhPhuNguyen1) September 30, 2024

[9] AI Glasses empower the sight impaired

AI Glasses, like the Meta Ray-Bans are available from John Lewis. The cost is about ÂŁ300. As seen below, they may be a great solution for the sight impaired, alongside a smartphone.

This is the type of stuff that makes me love AI pic.twitter.com/a5qtwpskuZ

— non aesthetic things (@PicturesFoIder) September 30, 2024

[10] Blindsight - Neuralink for the sight impaired

Maybe soon nobody will need AI Glasses to aid sight because their implant has given them super-powers!

The Blindsight device from Neuralink will enable even those who have lost both eyes and their optic nerve to see.

Provided the visual cortex is intact, it will even enable those who have been blind from birth to see for the first time.

To set expectations correctly, the vision… https://t.co/MYLHNcPrw6 pic.twitter.com/RAenDpd3fx

— Elon Musk (@elonmusk) September 17, 2024

[11] ChatGPT adds Canvas - like Claude with Artefacts

For the first time we are fundamentally changing how humans can collaborate with ChatGPT since it launched two years ago.

We’re introducing canvas, a new interface for working with ChatGPT on writing and coding projects that go beyond simple chat.

Product and model features:… pic.twitter.com/ruVvtCNKrV

— Karina Nguyen (@karinanguyen_) October 3, 2024

[12] The next wave of AI - and the nuclear power revival

Nvidia partnering with Accenture. presumably competing with Palantir in bringing AI to large business process.

Listen carefully to Jensen explaining the next and the biggest wave of AI.

This is a huge deal. $NVDA pic.twitter.com/G3aaMRwgv3

— The AI Investor (@The_AI_Investor) October 2, 2024

Google's clean power plans challenged by huge AI demands - to be helped by SMRs Small Modular Reactors

Alphabet CEO Sundar Pichai says Google are scaling up their compute infrastructure and working on 1 gigawatt+ data centers, while exploring options for powering them including small modular nuclear reactors pic.twitter.com/smBUiruuYg

— Tsarathustra (@tsarnick) September 21, 2024

"These aren't your parents' reactors" - US Dept of Energy

“These aren’t your parent’s reactors.”   @ENERGY’s Mike Goff breaks down 3 things you need to know about advanced reactors. 👇 pic.twitter.com/i0O13D4HPO

— Office of Nuclear Energy (@GovNuclear) September 25, 2024

[14] AI Swarms and Swarms of Drones

Emad Mostaque says that unlike human intelligence, AI can scale seamlessly, such that they produce swarms of intelligent agents that will improve one another by checking each other's work pic.twitter.com/w4Jtz1oQ72

— Tsarathustra (@tsarnick) September 29, 2024

Ten thousand drones in China - beautiful when used peacefully

đź’ĄRecord breaking drone show in China with +10,000 drones.

This tech will be weaponised.

Imagine 10,000 ai guided bombs swarming the sky. 🤨 pic.twitter.com/HxyLocrzy9

— Radar🚨 (@RadarHits) September 28, 2024

[15] AI and two Nobel Prizes

Probably not the last time that AI will be central to Nobel Prize awards. The only question is whether we will award prizes in future years to super-intelligent AIs who are independent of any particular humans ...

The 2024 #NobelPrize laureates in chemistry Demis Hassabis and John Jumper have successfully utilised artificial intelligence to predict the structure of almost all known proteins.

In 2020, Hassabis and Jumper presented an AI model called AlphaFold2. With its help, they have… pic.twitter.com/Izgx4FT23K

— The Nobel Prize (@NobelPrize) October 9, 2024