State of AI Report - Summarised

Key takeaways from the report, voice-diagnosis and more

Hi friends, this is Ben’s Bites, walking you through the sunny side of AI, avoiding the waffle and getting straight to the meat (or vegan alternative).

Today we’ve got:

  • State of AI Report

  • Diagnosing patients based on your voice

  • Product Roundup

  • AI images of the day

The State of AI Report

Nathan Benaich and Ian Hogarth (AI investors) put together a mammoth 113 slide report all about AI.

We read it, so you don't need to. Here's the key takeaways you need to know:

1. Speed & Capability

Research collectives have open-sourced breakthrough AI language, text-to-image and protein (not protein powder) models at a rapid pace. The pace is so fast that text-to-video capabilities are being released within days of each other.

But these are from Meta and Google. Open-source communities are surely soon to follow.

AI art has taken the internet by storm. Technology that was in the hands of few, is now in the hands of everyone with a smartphone/laptop.

Nvidia still holds the top spot for compute power. They’re used 90x more than 5 major AI chip startups combined.

Size really does matter! (with computers at least)

2. Breakthroughs

LLM (Large Language Models) can learn the language of proteins and be used for their generation and structure prediction.

Also, they learnt about Covid (sorry I mentioned it) which was used to predict high-risk Covid-19 variants far before the WHO designated them as variants.

How? The model ingests the protein and predicts its immune escape and fitness. So basically, AI can understand chemicals present in plants and can help discover new drugs and predict their properties. AI-first drug discovery companies have made 18 assets that are in active clinical trials today. Up from 0 only 2 years ago.

And to top off how AI is revolutionizing science, it can learn to control nuclear fusion reactors.

LLMs are also learning to use software tools. They can be trained to interact with search engines, web apps, and other software. Improved by watching us use them 👀.

3. Companies & Funding.

$48 Billion. That's how much funding has gone into AI-focused startups so far this year. It's expected to reach ~$71Bn.

Attention = Money. More students and researchers want to start companies based on their research.

4. Safety.

People are being vocal about their concerns for AI and the number of researches dedicated to AI safety has tripled.

Safety research has pulled in more money than in years past.

You can read the full report here.

Diagnosing patients based off your voice

What would you make with $100M?

Well, The National Institutes of Health is collecting 30k voices with medical disorders.

And with that data, they claim to be able to diagnose; neurological, voice, mood, respiratory and pediatric disorders like autism and speech delays.

Then they’ll turn that into an app which can advise you to seek medical attention.

But it’ll take 4 years.

You can read the post here.

Product Roundup

Difficulty coming up with stories for your kid? Once upon a bot has got you covered. Generate a random story about anything 

Elad Gil is hosting an interview with OpenAI co-founder, Sam Altman tonight (too late for me). I’ll post a summary of the interview tomorrow. 

No more boring backgrounds. Magically generate an infinite number of unique backgrounds 

Invoke AI - A version of Stable Diffusion you can run on your computer. It provides a polished Web interface and an easy-to-use command-line interface.

But either way, you need to be fairly technical to get it set up on your own computer.

AI Images of the day

I hope it's like GTA Vice City!

That's it! See ya ✌️

What did you think of this? Reply and let me know.

Join the conversation

or to participate.