- Riding the Wave
- Posts
- π Sam is back, OpenAI saved
π Sam is back, OpenAI saved
Claude 2.1 can read 500 page books now, ChatGPT voice rolled out to free users
Hola Surfersπ!
The OpenAI squad has regrouped, with a new board and Sam Altman back in the hot seat. Now that the dust is settling on all that drama, we can talk about all the other cool things happening in the space as well.
Hereβs everything that happened yesterday:
THE NEWS
π¨βπΌSam Altman is back as CEO of OpenAI
After a 5 day drama that sent OpenAI to the verge of collapse, Sam Altman is back in the driver's seat as CEO. The board had been in discussions with Sam yesterday with Emmett Shear (you know, the Twitch guy who was temporarily in charge) playing mediator.
A new, βinitial,β board of directors had been appointed, including former Salesforce co-CEO Bret Taylor, former US Secretary of the Treasury Larry Summers, and Adam DβAngelo, from the old board. Everyone on X has been side-eyeing Adam because of his ties to Poe, a rival to ChatGPT. But surprise, surprise, he's still on the board.
The big turnaround probably happened because the OpenAI team stood together and threatened to jump ship to Microsoft. That would've handed over the most capable AI team to one of the biggest industry players.
The board's keeping mum about why they axed Altman in the first place, though. A thing we might only find out from a memoir ten years down the line.
Now, the 'initial' boardβs job is to pick a bigger, 9-person board to sort out OpenAI's management mess. Microsoft, which has poured $10 billion into OpenAI, wants a seat at that table, and so does Altman.
But the important thing is that the team behind the industry leading technology stayed together and will continue to research and develop AI tools. Hopefully keeping true to their mission of building a safe AI for everyoneβs benefit.
we are so back
β Greg Brockman (@gdb)
7:41 AM β’ Nov 22, 2023
ποΈChatGPT Voice rolled out for all free users
Amidst all the management chaos, the team didn't lose focus and rolled out their voice feature to all free users. I've previously covered everything you need to know about this awesome feature right here.
In essence it is an all-knowing friend you can ask for advice. Can't remember how to play a game? Just ask. Need to practice for a job interview? It's ready to play it out with you. Stuck on a tough topic? Consider it your personal tutor.
You can interrupt it and ask followup questions. It never gets mad. Plus, it talks in pretty much every language, though some come with a thick American accent.
ChatGPT with voice is now available to all free users. Download the app on your phone and tap the headphones icon to start a conversation.
Sound on π
β OpenAI (@OpenAI)
8:43 PM β’ Nov 21, 2023
To try it out download the ChatGPT App. Free users will be able to use it with the GPT 3.5 engine.
πClaude 2.1 can read 500 page books now
While OpenAI is dealing with its internal issues, Anthropic isn't just sitting back. They've rolled out Claude 2.1, an upgrade to their AI model that's stepping up to challenge OpenAI's GPT series.
This new version of Claude boasts a whopping 200,000-token context window, outdoing GPT's 128,000. Thatβs roughly 500 double-spaced pages, or a really robust book. This means it can handle big tasks, like analyzing full codebases or digesting long literary works.
Although GPT-4 is still ahead in some areas, like code generation, Claude 2.1 has its own strengths and weaknesses. It's also gotten smarter, with fewer errors and a better sense of when it's out of its depth, choosing to play it safe rather than give wrong answers. Not everyone is pleased with this direction though.
The new version of Claude, 2.1, continues to be very good, though not yet close to GPT-4.
Even though it can now take in more info (400 pages!) it actually seems to have a bit more trouble with documents than Claude 2, missing a reference that the early one found & arguing more
β Ethan Mollick (@emollick)
6:23 PM β’ Nov 21, 2023
A standout feature of Claude 2.1 is its ability to use external tools. It can now interact with web interfaces, use calculators, access APIs, or perform web searches. This is particularly handy for complex questions, like recommending the best tech products.
This update is a significant move for Anthropic, showing they're a serious player in the AI field. With OpenAI distracted, Anthropic is making the most of the opportunity to advance.
ONE MORE THING
Stable Video Diffusion is here
The industry leader in open-source models, Stability AI, has released their first text-to-video model. The company behind the popular image generator, Stable Diffusion, is introducing a new tool for filmmakers called Stable Video Diffusion.
They have also announced an upcoming 'web experience' that features a Text-To-Video interface. You can access the model and sign up for the waitlist here.
β If you have one more minute
πͺA growing number of tech execs think AI is giving Big Tech βinordinateβ power
π¨How AI is expanding art history by identifying disputed artworks and reconstructing lost masterpieces
πFTC streamlines investigations of cases where AI is used to break the law
πOpenAI, Microsoft hit with new author copyright lawsuit over AI training
AI Art of the Day π¨
Hereβs a first test of Stable Video Diffusion. Not bad at all. However the background consistency will have to be addressed.
πππππππππππππππππππππππ
That's all for today, folks!
If you enjoyed this, please share this hand-crafted newsletter with a friend. It would make this writer's day!