Thoughts to text or speech prosthesis "is close at hand"!

Buried under the coverage of OpenAI & Google going head to head, far more interesting stories are lurking. But yeah, Google really f/cked that up.

Feb 14, 2023

Welcome to RiD! My side / passion project on all things language AI. It’s basically where you find out today, whats happening tomorrow (hopefully).

Here is the TL;DR list of this weeks newsletter:

Ummm…thoughts to text. Stanford making HUGE progress.
US Congress AI Caucus gains relevancy
Google really did f/ck that up
Here come the deepfakes
Nature magazine taking a productive stand on LLMs

Ummm…thoughts to text. Stanford making HUGE progress.

A Stanford University team claim to have achieved new state-of-the-art performance at 62 words per minute decoded speech, translating the thoughts of a patient into text on a screen. I’m just going to quote from the article. What a world we live in!

This state-of-the-art performance was achieved using two machine learning models, one to translate the brain’s electrical signals into predicted phonemes, the other to output a given word or sentence based on these predictions.
The paper further claims a 9.1% word error rate on a 50 word vocabulary, and an astonishingly low 23.8% word error rate on a 150,000 word vocabulary.

"The success of large language models over the last few years makes me think that a speech prosthesis is close at hand"

It’s a kind of ChatGPT UI with the world, for people suffering from paralysis or some other debilitating physical issue.

US Congress AI Caucus gains relevancy

“We’ve all been stunned by ChatGPT,” Rep. Don Beyer, D-Va., a 72 year-old member who went back to school recently to get a master’s degree in machine learning. “The fact that it's here now has really surprised me and I'm probably a little better informed than the average.” Read the rest on Semafor.

While commendable, thankfully congress and specifically the 32-member AI Caucus, is bringing in some experts from the field to help them get up to speed. What may have seemed like a sleepy obscure committee when formed (~2 years ago?), will now offer its members a chance to step into the spotlight, if they so desire. And they are politicians so…that was a rhetorical statement.

While the initial remit was “to make AI education, training, and R&D available for communities across the country”, ChatGPT seems to have stirred the hornets nest and now the discussion is much more around AI regulation and national security.

The European Parliament itself is expected to vote next month on what they call their AI Act, with final passage expected at the end of the year. The EU seems much further down this path, as they were with getting a handle on data rights with GDPR.

Whether governments can keep up with the pace of AI seems questionable, by which I mean the frontiers of whats possible and available in the world is both constantly evolving and being made available. That doesn’t mean however that there shouldn’t be some mechanism for collective agreement on the rules of the road and sorry web3 enthusiasts but we still need some deference to democratic and legal processes, even if they are in react than lead mode.

Google really did f/ck that up

Come on Google…

Shares of Google’s parent Alphabet closed down more than 7% on Wednesday after the company held an event that promoted its new artificial intelligence chatbot called Bard, one day after competitor Microsoft held its own event to show off new AI technologies in its competing search engine, Bing.

Seriously. As mentioned in the HardFork pod link (listen below), this seems like a hiccup to me and something they can get over but others disagree.

Scott Galloway @profgalloway

And overnight, it feels as if @Google has been poorly managed for a decade.

Regardless, what a shitty level of execution, given Microsoft is going full tilt into their OpenAI relationship and GPT integration into their suite of products. They clearly don’t think MS is wrong to do so but they look like they are reacting rather than prioritizing, if indeed they have been caught napping, even though they have all the research and resources to knock this out of the park. Is this just a clear case of the incumbency trap?

Anyway, they are making Satya look like a playa who can match words with action…

Ethan Caballero @ethanCaballero

"This new Bing will make Google come out and dance, and I want people to know that we made them dance." - @satyanadella

Here come the deepfakes

The influencer / doctor / researcher Dr. Andrew Huberman, along with Joe Rogan have beeb deepfaked on a supplement ad on tiktok. The sound works better than the video but at first glance all looks pretty kosher.

Jimmy Farley @JimmyFarley00

My first time seeing a deepfake ad in the wild on TikTok… 😳 How tf is this legal?

Over on 4chan, the playground of the disgruntled, people are also iterating on celebrities saying whatever they want them too. 11Labs, mentioned various times here, is gaining traction, given the competence of their platform and started to ramp up marketing.

Vid cred - G16.

Nature magazine taking a productive stand on LLMs

From Nature magazine:

Conversational AI is likely to revolutionize research practices and publishing, creating both opportunities and concerns. It might accelerate the innovation process, shorten time-to-publication and, by helping people to write fluently, make science more equitable and increase the diversity of scientific perspectives. However, it could also degrade the quality and transparency of research and fundamentally alter our autonomy as human researchers. ChatGPT and other LLMs produce text that is convincing, but often wrong, so their use can distort scientific facts and spread misinformation.
We think that the use of this technology is inevitable, therefore, banning it will not work. It is imperative that the research community engage in a debate about the implications of this potentially disruptive technology.

This is the only path forward, be it in higher ed, middle and high school and across industry.

»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»