Artificial Intelligence

A new report from The Washington Post says that senior OpenAI staffers had indicated to the board Altman had been “psychologically abusive.” The board had also worried that it couldn’t keep Altman accountable.

All of that contributed to Altman’s firing, the report says — though, as we now know, Altman ultimately came out on top.

Project Ellmann, named after biographer and literary critic Richard David Ellmann, will take users’ search results and photos to make a chatbot able to answer “previously impossible questions.”

CNBC saw documents presenting Ellmann and said the goal is to present a “bird’s-eye” approach to a person’s life. Ellmann allegedly categorizes moments into chapters: for example, the “college” chapter or the “becomes a parent” chapter. People could use Ellmann to ask questions like “did I have a pet,” or “when did my sibling last visit.”
Google tells The Verge that Ellmann was an early internal experiment. “Google Photos has always used AI to help people search their photos and videos,” the company said. It added if they decide to roll out features like Ellmann they would “take the time needed to ensure they were helpful and designed to protect users’ privacy.”

Personally, this type of hyperpersonalization is lost on me because I have a good enough memory of my life that I wish I could forget certain moments.

The CMA is asking whether the relationship and “recent developments” between the two companies constitutes a merger that impacts competition in the UK.

“The only thing that has changed is that Microsoft will now have a non-voting observer on OpenAI’s Board, which is very different from an acquisition such as Google’s purchase of DeepMind in the UK,” said Microsoft’s president Brad Smith to Bloomberg.

Moises, the audio tech app behind it, officially launched Music.AI today. It offers a mix of proprietary AI models and third-party tools for commercial use. The APIs also include one for vocal extraction, downbeat detection, and a tool that can transform a song into a video (complete with synchronized lyrics) — among others.

Helen Toner, one of the board members who fired Sam Altman and then ousted herself when Altman returned, has given an interview to The Wall Street Journal. She is unable to give a specific reason for Altman’s firing beyond “Our goal in firing Sam was to strengthen OpenAI and make it more able to achieve its mission.”

Listen, babe, I am fully happy to hear you out, but if being full of shit is a firing offense for CEOs, everyone in the Fortune 500 would be looking for a new job.

Google launched its Gemini AI model earlier today to compete with OpenAI’s popular GPT-4. While Google has a number of videos demonstrating Gemini’s capabilities, the one below stood out to me.

This multimodal AI model is capable of reasoning across images, audio, video, code, and, of course, text. So you can start drawing, and Gemini will understand you’re drawing a duck, or you can set some cups down on a table with a paper ball and Gemini reasons you want to play a game. If Gemini can understand my poor attempts at doodling, then I’ll be the one shouting, “what the quack!”

And wow, I do not have them. So I flew to Berlin to learn from a “Promptographer,” and I came home with more generated pancakes than I know what to do with.

The phrase became an internal joke after ChatGPT’s popularity exploded right out of the gate, according to the NYT’s recap of its launch a year ago and the reaction among Big Tech companies.

Google and Meta scrambled AI teams to launch competing products — even if that meant removing some guardrails — like Bard and LLaMa. And Microsoft’s rush to beat Google had Satya Nadella saying, “We have a big order coming to you, a really big order coming to you,” to Nvidia’s Jensen Huang as he ordered $2 billion in chips.

This chart from Omdia Research estimating Nvidia’s largest customers this year has been making the rounds in my social media feeds.

As I wrote in an earlier issue of Command Line, these H100s are essentially the tech industry’s new gold, since they are the preferred workhorse for powering generative AI. The gap in shipment volume between Meta, Microsoft and everyone else is quite something, and tracks with what I’ve heard from sources in recent months.

Nothing important happened while I was gone, right?

The answer, according to a pre-print study, is about one. Researchers at AI company Hugging Face and Carnegie Mellon University found that general-purpose AI models like GPT-4 are “orders of magnitude” more power-hungry than purpose-made models powering products like Google Translate.

The study, though not yet peer-reviewed, puts into context the environmental cost of generative AI, particularly of inefficient models (one image from the least efficient image-creating model can use as much CO2 as an average gas car driving about 4 miles, for instance).