🤖 This Week’s AI Insight 🤖

Meta announces the ultimate AI object detector, which you can try now for free

Foundation models like GPT-4 powering tools like ChatGPT are the reason you’re reading this newsletter now.

The paradigm shift they’ve caused in AI has transformed the sector from one with great potential to one that’s shaping not only how we work, but ultimately how we live our lives.

Until now, this shift was only visible in Generative AI, the AI field that generates text like ChatGPT, images like Stable Diffusion, and soon video, while other applications remained wishful thinking.

But all of a sudden, Meta has released what we can consider the first foundation model for image segmentation, and the results are breathtaking.

And more importantly, you can try it right now from your own browser.

Computer Vision on Steroids

Meta research has released its Segment Anything Model, or SAM.

This model, trained with a huge dataset of 11 million images with up to a billion masks (a mask being the detected object in an image), shows incredible promise to be used as a generalistic model for object detection or generation.

In layman’s terms, given an image and a prompt (an instruction to what you want to detect, given as a cursor point, a box, or even free text) the model identifies the object you’re aiming for.

And as you can see in the above image sampled from the model, SAM is capable of identifying up to 100 different objects per image with uncanny accuracy.

But how does this model work and how can you try it?

Transformers as a way of living

SAM relies heavily on Transformers, the architecture used by ChatGPT that, by introducing the concepts of positional encoding, multi-head attention, and parallelizable computation, allows machines to extract context from data (text or images) incredibly well.

SAM works as follows:

• A Vision Transformer (ViT) generates an embedding of the image (a representation of the image in the form of a vector)

• A prompt encoder (they used OpenAI’s CLIP encoder) obtains the embedding of the prompt

• A mask decoder then combines the information from these two embeddings and maps it into its output, a mask that represents the detected object in the image.

Et voilà.

Bottom line, SAM can represent the next big leap for AI, taking computer vision to the same levels of hype that Generative AI solutions are enjoying today.

Luckily for you, you don’t have to trust Meta or me to understand how amazing this tool is, you can try it for yourself and get mindblown by giving it your own images.

Enjoy!

👾AI news for the week👾

🤔 Thought-provoking article to why LLMs like GPT-4 “make stuff up”, and ways to mitigate it.

♾ Google announces new supercomputer, says it beats the industry-leading NVIDIA chips

📣 Meta to launch Generative AI Ad product

👩🏻‍⚖️ Can AI be sentenced for defamation? We’ll soon find out

❤️ Google’s ex-CEO warns that people will fall in love with AI systems

😨 AI-generated video is scary as hell

🔐 This Week’s Crypto Insight 🔐

How zk-proofs converge AI and Blockchain

While many investors claim Crypto is dead while they pivot toward AI, many of them don’t realize that AI actually needs Crypto.

But how is that?

Well, it’s thanks to zk-proofs, according to one of the leading venture capital firms in the world, a16z.

Running trusted computations

As some AI labs like OpenAI have decided to take their models private, making them only accessible via API, this means you really can’t see what’s going on behind this interaction.

Naturally, you will tend to believe that OpenAI is being honest and won’t lie to you, but they could and it would be impossible for you to verify.

As OpenAI handles the costs of running these models, which are huge, this forced them to make GPT-4 only accessible by paying a monthly fee.

But what would prevent OpenAI from claiming you’re using their GPT-4 model while, in reality, running the 3.5 version while making you pay for the most expensive GPT-4?

Well, here is where zk-proofs come into play.

A succinct marvel

Zk-proofs allow you to prove a statement is true without actually showing the statement.

In layman’s terms, taking the previous example, you could get high assurance that you’re running GPT-4 through the API without forcing OpenAI to actually show you the model’s computation.

The process is simple.

The prover generates a series of small, succinct proofs that, combined, make it statistically very improbable that the statement the verifier is trying to verify - that you’re indeed running GPT-4 in the previous case - is false.

This can also be applied to things like data to verify that yours is not being used for the training of a model, or basically any privacy-preserving case you can think of.

Sounds awesome!

But where do blockchains fall in all of this?

Protecting integrity

As always, there’s only one thing you need to know about blockchains.

By being decentralized, they protect data integrity, making them a trustworthy source of truth without the need for external trust enablers.

Thus, assuming zk-proofs will be crucial to ensure a privacy-preserving environment for AI, blockchains will be extremely necessary for storing the actual proof.

If you don’t use a blockchain, zk-proofs lose a great deal of credibility, as without blockhains they could be easily tampered with.

Thus, blockchains guarantee a trustless environment where all parties simply trust the power of cryptography and statistics to work; an environment we can all trust and feel secure in.

Read more on a16z’s blog post.

📉Crypto news for the week📈

🗽 A Bitcoin Ordinals megathread

🤫 Macs from 2018 onwards have the Bitcoin whitepaper in them

🪙 Bitcoin’s case for +$300,000 value, Schröndiger’s Bitcoin theory

🦜 Twitter changed its logo to Doge, then removed it again

🤩 “What people call intelligence boils down to curiosity”

While we debate if GPT-4 is sentient, we’re one step closer to reading an Octopus’ mind.