Nobody Cares: Technology-only Edition

[Free]

Microsoft barrels ahead with AI plans, opens up Bing Chat preview to everyone

Microsoft also plans third-party plugins, more images, and other enhancements.

arstechnica.com

Today, Microsoft continued its push to integrate generative AI features into all of its products, announcing a range of small updates to its ChatGPT-powered Bing chatbot and moving the product from a closed, invite-only preview to an open preview that anyone can join. This next phase of the "new Bing" comes just three months after Microsoft introduced it and further demonstrates Microsoft's commitment to putting AI features in all of its products as quickly as possible.

The "Open Preview" phase of the New Bing makes the chatbot and other features available to anyone with a Microsoft Account, and it coincides with a push to make the chatbot "more visual." Bing Chat will soon be able to generate charts and graphs as part of its answers, and the Bing Image Creator will be available in all of the 100-plus languages supported by Bing. A new visual search feature will let you search for images by uploading a similar image to Bing Chat first.

Microsoft is also planning to expand the amount of history and context Bing Chat can remember, "moving from single-use chat/search sessions to multi-session productivity experiences with chat history and persistent chats within Edge," writes Microsoft VP Yusuf Mehdi. This means you'll be able to return to previous chat sessions, chats can stay open while you browse in Edge, and chats will become "more personalized" as Bing Chat "bring context from a previous chat into new conversations." The latter feature will be rolled out "over time" since it's something Microsoft is still "exploring."

 

[WolfEyes]

Yay? I barely use Bing as it is, now I have even more reason to not use it at all.

Needs more cowbell.

 

[Dakota Tebaldi]

Warning of AI’s danger, pioneer Geoffrey Hinton quits Google to speak freely

Most people thought it was way off. And I thought it was way off."

arstechnica.com

Some critics have cast a skeptical eye on Hinton's resignation and regrets. In response to The New York Times piece, Dr. Sasha Luccioni of Hugging Face tweeted, "People are referring to this to mean: look, AI is becoming so dangerous, even its pioneers are quitting. I see it as: The people who have caused the problem are now jumping ship."

I mean, yeah. I don't know if I would quite put it that way - that it's guilt over "causing a problem", I don't see that in his words; but "I quit Google so I can express the same very generic and nonspecific concerns about AI in general that people have been expressing since like the 80's" is quite a bit less than momentous honestly?

 

[Khamon]

Perhaps, to him, it sounds better than retiring from a glittering career and beginning a new one. Political activism is a distinctly new field of study for him. I hope he enjoys it as much as he did working in Computer Science.

 

[Bartholomew Gallacher]

An internal Google document has been leaked, topic: "We Have No Moat, And Neither Does OpenAI".

We’ve done a lot of looking over our shoulders at OpenAI. Who will cross the next milestone? What will the next move be?
But the uncomfortable truth is, we aren’t positioned to win this arms race and neither is OpenAI. While we’ve been squabbling, a third faction has been quietly eating our lunch.
I’m talking, of course, about open source. Plainly put, they are lapping us. Things we consider “major open problems” are solved and in people’s hands today. Just to name a few:

• LLMs on a Phone: People are running foundation models on a Pixel 6 at 5 tokens / sec.
• Scalable Personal AI: You can finetune a personalized AI on your laptop in an evening.
• Responsible Release: This one isn’t “solved” so much as “obviated”. There are entire websites full of art models with no restrictions whatsoever, and text is not far behind.
• Multimodality: The current multimodal ScienceQA SOTA was trained in an hour.

While our models still hold a slight edge in terms of quality, the gap is closing astonishingly quickly. Open-source models are faster, more customizable, more private, and pound-for-pound more capable. They are doing things with $100 and 13B params that we struggle with at $10M and 540B. And they are doing so in weeks, not months. This has profound implications for us:

• We have no secret sauce. Our best hope is to learn from and collaborate with what others are doing outside Google. We should prioritize enabling 3P integrations.
• People will not pay for a restricted model when free, unrestricted alternatives are comparable in quality. We should consider where our value add really is.
• Giant models are slowing us down. In the long run, the best models are the ones
  which can be iterated upon quickly. We should make small variants more than an afterthought, now that we know what is possible in the <20B parameter regime.

In other words: since mid of March the Llama weights have been leaked, Opensource activists have made technical progress and created innovations in an unmatched speed, and solved the problem of scalability which the tech companies were unable to tackle. Opensource is destroying right now the business model of the AI tech giants like Google and OpenAI.

Google "We Have No Moat, And Neither Does OpenAI"

Leaked Internal Google Document Claims Open Source AI Will Outcompete Google and OpenAI

www.semianalysis.com

 

[Argent Stonecutter]

An internal Google document has been leaked, topic: "We Have No Moat, And Neither Does OpenAI".

"Open-source models are [...] more private"

That's the one that scares Google, I suspect.

 

[Noodles]

On that Google thing.

Is there a text based LLM you can run locally yet? I have Stable Diffusion going but I would love to have a local text one to mess with.

 

[Sid]

I can't imagine anyone with a Mac not already using Time Machine.

You mean a Big Mac or a cheese burger?

 

[Free]

You mean a Big Mac or a cheese burger?

Cheeseburgers come with Time Machine?

 

[Sid]

Cheeseburgers come with Time Machine?

They used to have all kinds of toy gadgets with a happy meal, right?
Maybe the cardboard burger sellers changed their policys, I don't know. I prefer local fast food places.

 

[Free]

I don't know, I haven't been in a McD for ages.

Don't you mean a MacD?

 

[Noodles]

Cheeseburgers come with Time Machine?

> Take a bite

> Roll to back up

> Bite is restored

> Infinite food.

Food scarcity hates this one weird trick!

 

[Free]

> Take a bite

> Roll to back up

> Bite is restored

> Infinite food.

Food scarcity hates this one weird trick!

If only we had infinite cheeseburgers.

 

[Free]

The new BLOMFU keyboard layout.

 

[Bartholomew Gallacher]

OpenLLaMA has been launched, which is an open-source recreation of Facebook's LLama large language model with help of scientists from Berkeley. Commecial use is permitted, and it uses way less parameters (around 7 billion) than Facebook's Llama (around 300 billion), which means it can on more consumer grade hardware.

The goal is to bring it up to par with Facebook's language model over time.

https://the-decoder.com/openllama-is-a-fully-open-source-llm-now-ready-for-business/

 

[Free]

 

[Argent Stonecutter]

It's not rocket science, it's rocket comedy.

 

[WolfEyes]

If they really were "pulling it off" how come that last launch was such a complete and utter failure?

Rhetorical question.

 

[Argent Stonecutter]

They're pulling off the trick of being uncool.

One of them has a rocket that looks like something from a 50s soft-porn SF book cover, and the other isn't satisfied with just making really nice conventional rockets and is obsessed with 50s Mars Colony SF.

 

[Casey Pelous]

If they really were "pulling it off" how come that last launch was such a complete and utter failure?

Rhetorical question.

Probably because they were "pulling it off!"



Thank you, thank you. I'm here till Thursday. ...