Nobody cares about "AI" (Chatbot: I disagree.)

[Bartholomew Gallacher]

Background is that many people considered the voice very similar to Scarlett Johansson. This is OpenAI telling us: no, it was another voice actor.

 

[Argent Stonecutter]

... who happened to sound by sheer coincidence just like Scarlett Johansson.

 

[Dakota Tebaldi]

The excuse might be believable if OpenAI hadn't specifically asked Johansson, twice - including two days before the launch - if they could use her voice.

But that's how it went down. Altman asked for permission and was refused, so in the usual techbro fashion he just did it anyway.

 

[Dakota Tebaldi]

Art portfolio site DeviantArt has effectively been overrun; monetization features and official social media posts meant to surface and highlight the work of popular artists on the site now just showcase bot accounts selling AI art to each other in order to boost their own sales footprint. The siterunners appear to be uninterested in addressing the problem, since they don't see it as a problem.

 

[Bartholomew Gallacher]

BTW ChatGPT names Sarah Elmaleh as the voice of Sky by OpenAI.

 

[Noodles]

... who happened to sound by sheer coincidence just like Scarlett Johansson.

Leo on the TWIT Podcast has been using it a lot on the show for demos. When I first heard it, I had assumed he had trained it to sound like ScarJo because he has been saying he wishes he had "Her" from the movie for a while now. Specifically ScarJo from "Her."

The sound alike is definitely not a coincidence.

 

[Free]

Hmm.

Alibaba Sparks China AI Price War With Spate of Steep Discounts

Alibaba Group Holding Ltd. slashed prices for a clutch of artificial intelligence services by as much as 97%, spurring an immediate response from Baidu Inc. in potentially the start of a price war in China’s nascent AI market.

Baidu Cloud said it would offer free services based on its Ernie AI models on Tuesday, hours after Alibaba offered deals on nine products built atop its own Tongyi Qianwen. ByteDance Ltd. last week announced pricing for AI services that it said were 99% lower than Chinese industry norms, using Ernie and Alibaba’s Qwen as benchmarks.

The tit-for-tat maneuvers mark the opening salvos of a price-based battle within AI, a field that’s attracting billions of dollars of investment from startups and internet leaders including Tencent Holdings Ltd. The flurry of investment has created scores of AI models and spawned many more consumer and enterprise products in turn, all fighting for the critical mass of users needed to speed AI development.

Discount AI services. I'm sure this will turn out well.

 

[Free]

Going forward...AI everywhere in the newsroom. That doesn't sound ominous, does it...

 

[Free]

Here’s what’s really going on inside an LLM’s neural network

Anthropic's conceptual mapping helps explain why LLMs behave the way they do.

arstechnica.com

With most computer programs—even complex ones—you can meticulously trace through the code and memory usage to figure out why that program generates any specific behavior or output. That's generally not true in the field of generative AI, where the non-interpretable neural networks underlying these models make it hard for even experts to figure out precisely why they often confabulate information, for instance.

Now, new research from Anthropic offers a new window into what's going on inside the Claude LLM's "black box." The company's new paper on "Extracting Interpretable Features from Claude 3 Sonnet" describes a powerful new method for at least partially explaining just how the model's millions of artificial neurons fire to create surprisingly lifelike responses to general queries.

An interesting read.

 

[Free]

Too many serious posts. Gotta switch things up...

 

[Free]

The flood of AI-specific news continues.

Political ads could require AI-generated content disclosures soon

This follows an FCC crackdown on AI-generated voices in robocalls.

www.theverge.com

 

[Bartholomew Gallacher]

GitHub Copilot gained a set of extensions, including an official McKinsey&Partner extension. Yeah sure, when I do think GitHub McKinsey comes on my mind... NOT!

Introducing GitHub Copilot Extensions: Unlocking unlimited possibilities with our ecosystem of partners

The world of Copilot is getting bigger, improving the developer experience by keeping developers in the flow longer and allowing them to do more in natural language.

github.blog

 

[Khamon]

Here’s what’s really going on inside an LLM’s neural network

Anthropic's conceptual mapping helps explain why LLMs behave the way they do.

arstechnica.com

It's a fascinating challenge to normalize a system that operates using fuzzy many-to-many relationships. One might pursue this avenue in one's advanced age and become Khamon Calvin Fate. It will be a suitable end to a glittering career.

 

[Noodles]

The flood of AI-specific news continues.

Political ads could require AI-generated content disclosures soon

This follows an FCC crackdown on AI-generated voices in robocalls.

www.theverge.com

Oh, I am sure the Russians will totally abide by that.

 

[Dakota Tebaldi]

Google's AI doesn't understand what "satire" is, obviously, so it just obliviously treats satirical "news" sites like The Onion as sources like any other news site.

 

[Free]

Google's AI doesn't understand what "satire" is

Should that be called the Star Trek Data problem?

 

[Free]

 

[Free]

Search may be about to re-up.

Google promised a better search experience — now it’s telling us to put glue on our pizza

Google’s new AI Overview feature is telling users a whole bunch of wrong answers.

www.theverge.com

Imagine this: you’ve carved out an evening to unwind and decide to make a homemade pizza. You assemble your pie, throw it in the oven, and are excited to start eating. But once you get ready to take a bite of your oily creation, you run into a problem — the cheese falls right off. Frustrated, you turn to Google for a solution.

“Add some glue,” Google answers. “Mix about 1/8 cup of Elmer’s glue in with the sauce. Non-toxic glue will work.”

So, yeah, don’t do that. As of writing this, though, that’s what Google’s new AI Overviews feature will tell you to do. The feature, while not triggered for every query, scans the web and drums up an AI-generated response. The answer received for the pizza glue query appears to be based on a comment from a user named “fucksmith” in a more than decade-old Reddit thread, and they’re clearly joking.

This is just one of many mistakes cropping up in the new feature that Google rolled out broadly this month. It also claims that former US President James Madison graduated from the University of Wisconsin not once but 21 times, that a dog has played in the NBA, NFL, and NHL, and that Batman is a cop.

What may be most amazing is the Google spokezombie's response.

Google spokesperson Meghann Farnsworth said the mistakes came from “generally very uncommon queries, and aren’t representative of most people’s experiences.”

I guess a lot of people don't search about pizza making. Or about dogs. Yeah.

 

[Dakota Tebaldi]

OpenAI has signed a content provision deal with Rupert Murdoch's News Corp for its ChatGPT engine.

Several news organizations have signed deals allowing OpenAI to use their publications for training data; OpenAI CEO Sam Altman has been aggressively reaching out to make such deals in the wake of being sued by The New York Times. But the deal with News Corp - owner of conservative publications like the New York Post and the Wall Street Journal, and sister company of Fox News - is unique, in that it establishes that ChatGPT will also directly source answers to users' news and current-events-related questions from News Corp's publications.

 

[Noodles]

Google's AI doesn't understand what "satire" is, obviously, so it just obliviously treats satirical "news" sites like The Onion as sources like any other news site.