For what it's worth, as always 99% benchmarks are very unreliable and per-task performance still greatly differs per model, with plenty of cases where results are wildly different.
I have a task I use in my work where Gemini 1.5-Pro is SOTA. Handily beating o1, Sonnet-3.5, Gemini-exp and everyone else, very consistently and significantly.
The newer/bigger models are better at reasoning and especially coding, but there's plenty of tasks that have little overlap with those skills.
* The article points out that reliable sources say the media org runs a network of fake news sites pushing BJP propaganda.
* The media org sues Wikipedia.
* The judge threatens to block Wikipedia in India, and demands the doxxing of the editors who made these observations about what sources say.
* Somebody starts a Wikipedia page about this civil case.
* That page now says "The Wikimedia Foundation has suspended access to this page due to an order by the Delhi High Court", but the one the case is about is still up.
Was there no such order about the Asian News International page? Or there was, but the WMF is ignoring that one while complying with this one?
I don't really get it, although "refrain from publishing information about an ongoing trial in case you prejudice the outcome" would be a reasonable request to comply with for ethical reasons. But they make it sound like they were forced to block this page and didn't want to. But not the page this page is about. Huh?
Edit: I think I see now, thanks to the above link about "On-wiki discussion". Something about the vagaries of law means blocking the meta-level article, but not the original one, is necessary if they want to appeal, years down the line when they get a chance. So it's strategic.
Does the community license let companies fine-tune it or retrain it for their use cases?
There are significant restrictions on it so it's not fully open-source, but maybe it's only a real problem for Google and OpenAI and Microsoft.
Open source has turned into a game of, what's the most commercial value I can retain, while still calling it open-source and benefiting from the trust and marketing value of the 'open source' branding.
But the nature of LLMs is stochastic, nothing is 100%. The LLM vendors aren't dummies and train hard for this use case. But you still need a prompt that OpenAI can handle, and validating / fixing the output with an output parser, and retrying.
In my experience asking for simple stuff, requesting json_object is reliable.
with LangChain even! eye-roll, you can't really title the post 'every way' and omit possibly the most popular way with a weak dig. I have literally no idea why they would omit it, it's just a thin wrapper over the LLM APIs and has a JSON output parser. Of course people do use LangChain in production, although there is merit to the idea of using it for research, trying different LLMs and patterns where LangChain makes it easy to try different things, and then using the underlying LLM directly in prod which will have a more stable API and fewer hinky layers.
this post is a little frustrating since it doesn't explain things that a dev would want to know, and omits the popular modules. the comment by resiros offers some good additional info.
it's a bust-out, you identify a patsy and you stick it to them.
sometimes the company is worth more dead than alive, the parts are worth more the whole, especially when you can leave someone holding the bag, and the PE company gets paid to make them dead.
in any event the company is worth more to an extremely unscrupulous buyer than as a going concern in public markets.
He's a UI strategist, from a UI perspective some of these AI things are a mess.
But Google says that their mission is to organize the world's information and make it universally accessible and useful. A technology that understands and generates human language, with all its idiosyncrasies and connotations, is probably pretty important to that mission. And OpenAI stole a march on Google in commercializing it.
Maybe they justifiably panicked a little because they were starting to miss the boat.
It's a fine line between running around with your head cut off, and entering terminal decline due to too little too late.
That mission of organizing the worlds info fell to the way side, after they became the world's largest Human Attention trading (some call it thieving) marketplace.
They can't reconcile both missions.
Ideally Alphabet should be running this as separate company that is free from any obligation to incorporate what they build, into Google's existing line up of garbage no customer ever asked for. That ever growing landfill has mostly been built up to increase digital real estate, on which to sell ads. And if it all burns down its good for the planet.
When I heard it, I thought, wow, they licensed Scarlett Johansson, what an amazing Easter egg.
If they didn't and just cloned her voice, it's more disregard for creators and artists than I would have thought possible. What were they thinking?
Edit after reading the official story... not sure I believe it, seems disingenuous, at best they chose someone because they really really sounded like Scarlett Johansson, and no one said, it might be a problem.
They cannot disclose the identity of the person involved in order to protect their privacy. This is a very convenient cloak, leaving no way to know if they cloned Scarlet Johansson's voice or if it's from someone else.
Fortunately for Scarlett, she can just sue them and force them to tell her. She also isn’t shy to litigate. I was surprised that Sam wanted to poke that particular beehive.
I know trademark litigation is wishy-washy, but can she even claim her voice is unique enough to claim some form of infringement? If I just happen to sound like her, am I a walking infringement?
Voices such as Shatner or Walken are at least as much about speech patterns than the voice, giving you another axis to compare voices against, so I can somewhat see those as being trademarkable. But when I hear the ChatGPT voice, it just sounds like "slightly-flirty generic female voice 03" to me.
Sam Altman and other OpenAI employees have been referencing the movie ahead of the presentation[1][2]. I think it'd be really challenging to prove that they didn't have that exact outcome in mind.
I’m not sure it’s a trademark question so much as an appropriation of likeness question. I just meant that they can’t really train on her actual voice samples and then conceal whether they did it; that’s what discovery is for.
OpenAI is going to do a very good job of pretending that it stealing literally everything on the internet and re-selling it is some unique new activity that's not happened before. That's their core business. What they're not going to be able to do is pretend that hiring a soundalike and then making repeated references to the original voice is ok. This isn't new legal ground no matter how much money you throw at it. Otherwise every advert on TV would have some no-name actor doing impressions of well known actors. They will have to pay ScarJo a lot of money, and probably they'll have to stop using the voice too - because they've pissed her off so much at this point.
reply