April 09, 2024

Ryan here. Am I the only one underwhelmed by the eclipse? Today's AI+ is 1,205 words, a 4.5-minute read.

Join me and Axios' Hope King in NYC tomorrow evening for an event featuring conversations with Moms First founder and CEO Reshma Saujani, Etsy CEO Josh Silverman and Greenhouse president and co-founder Jon Stross on the upskilling needed to develop an AI-equipped workforce. RSVP here.

1 big thing: AI startups dominate Enterprise Tech list

Illustration: Brendan Lynch/Axios

AI companies dominate this year's list of top startups focused on selling technology to businesses, according to the latest Enterprise Tech 30 list compiled by venture capital firm Wing and shared first with Axios, Ina reports.

Why it matters: Businesses have been investing heavily in generative AI over the past two years, though the jury's still out on how much the technology will boost productivity or profitability.

  • "This year's ET30 is heavily weighted towards AI, especially those companies building AI's infrastructure, tooling and core model technologies," says Peter Wagner, a founding partner at Wing.

By the numbers: The list includes 40 companies — 10 early-stage, 10 mid-stage, 10 late-stage and 10 "giga" companies.

  • Of the 40 companies, 17 are new to the ET30, and seven are in AI models.

Between the lines: The group of early-stage companies is the most dominated by AI companies, while the late-stage and "giga" companies are a wider mix.

  • Wagner notes that's unsurprising because the larger firms were founded well before the "GenAI moment" touched off by the release of ChatGPT in 2022.

Zoom in: The top early-stage company is LangChain, which makes tools for developing applications powered by large language models.

  • The top mid-stage company is Pinecone, a data platform that helps companies get better results from large language models through training on specific data — a technique known as RAG (retrieval augmented generation).
  • "The rest of these two cohorts are filled with other AI-first companies, demonstrating the industry's sharp shift in this direction," Wagner says.

Flashback: AI was the dominant theme in last year's list as well, with a third of companies focused on generative AI.

2. Google will "ground" AI in search for businesses

Google Cloud CEO Thomas Kurian. Photo: Cayce Clifford/Bloomberg via Getty Images

Google is launching a new effort today that will let businesses use results from its flagship search engine to help "ground" the results of generative AI queries, Ina reports.

Why it matters: The tendency of large language models to hallucinate, or make things up, is a key hurdle to broader business adoption of generative AI.

Driving the news: Google is announcing the "grounding" effort as part of a slew of announcements at its three-day Google Cloud Next conference, which kicks off today.

  • Google says that by integrating Google search into Vertex, its AI offering for businesses, it can give users access to more recent information and citations indicating the sources for key data.
  • Microsoft has also aimed to use its Bing search engine to help offer citations along with answers returned by its Copilot chatbot.
  • Google allows businesses to "ground" results from AI queries using their own data, though that effort is at an earlier stage. Google announced new connections to Workday and Salesforce data, with plans to add more options over time.

Zoom in: Among the early customers of Google's enterprise AI tools has been Verizon, which is using the technology for everything from deploying 5G networks to creating personalized marketing to revamping how it handles customer service.

  • In that last area, Verizon is using generative AI both to automate more queries and to help put the right information in front of the humans who are answering the tougher customer questions.
  • "We reduce the cognitive load on our agents and have them instead focus on building that emotional connection with the customers that is really the name of the game," Verizon chief information officer Shankar Arumugavelu tells Axios.

Separately, Google is using the Google Cloud Next event to introduce several other products for its Cloud business customers.

  • It's debuting Google Vids, a new AI-powered video editing app that will become part of Workspace, the company's paid productivity suite. Vids is set to be released to Google's experimental Workspace Labs in June.
  • There's also a new $10 per person per month AI-powered messaging and meeting service, as well as a similarly priced security service that can be added to Google Workspace.
  • Google is expanding its Imagen text-to-image engine, with Imagen 2.0 supporting the creation of animations in addition to still images.
  • The company is announcing Axion, its first custom datacenter chip with an Arm core, along with general availability for its most powerful homegrown Tensor processor, the v5p. Google is also committing to offer two of Nvidia's forthcoming Grace Blackwell chips early next year.

3. Gaming finds new role as AI's open ethics lab

Illustration: Lindsey Bailey/Axios

Gaming is a key testbed for AI technologies and the values society is embedding in digital worlds, a top gaming industry executive and AI researcher argues in a new book, Axios' Alison Snyder reports.

Why it matters: Gaming, AI and other tech industries continue to be dominated by men — even though plenty of women are avid gamers, and women are more likely than men to be exposed to AI in their work.

"When there's only one group of people deciding the rules and mechanics of a game, they impose their worldview on the rest of us. In that way, their games shape the communities in which we all live," writes Songyee Yoon in her new book, "Push Play: Gaming For a Better World."

  • Yoon is president and chief strategy officer at NCSOFT, one of the world's largest online game developers, headquartered in South Korea.

Zoom out: Yoon draws parallels between the biases perpetuated in games featuring male-only heroes and female characters sporting "impossible body shape[s] with no muscle and carrying big weapons" and the biases that get built into AI when it's trained on skewed data.

  • Studies have found people absorb those biases from AI.
  • Once biases and values are codified in a game or an AI model, "it becomes a part of our culture" and can become "the de facto view of our society," Yoon tells Axios.

The big picture: The gaming industry has long faced controversies over gender, representation and bias — from Gamergate 10 years ago to current attacks against supposedly "woke" developers.

  • The Anti-Defamation League found 76% of U.S. adult gamers in 2023 said they experienced harassment while playing online multiplayer games.

Yoon says there is mounting evidence that "avatars can have therapeutic effects for certain gamer populations."

  • She cites a study that found transgender and gender-nonconforming gamers "reported higher feelings of self-worth and confidence when they played avatars that matched their gender expression."
  • Games in that sense can provide a space for users "to explore, develop and rehearse their gender identities," Yoon argues.

The latest: NCSOFT last year announced it developed its own large language models for game production, including "digital humans, generative AI platforms, and conversational language models." Other game developers are deploying generative AI tools as well.

The bottom line: "There is so much that we still don't know about this new tool, and the more we play with it in our games, the more we can realize its potential," Yoon argues.

  • AI "provides a mirror for all of us," Yoon says.

4. Training data

  • Microsoft announced a new AI hub in London (the home of new Microsoft AI CEO Mustafa Suleyman) to be led by former Inflection and DeepMind scientist Jordan Hoffmann. (Reuters)
  • Paddy Cosgrave is once again CEO of Web Summit — he was forced out six months ago after criticizing Israel's response to Hamas' Oct. 7 attacks on the country. (RTE, Axios)
  • Trading places: John-Michael Murphy has joined the Perplexity.ai technical team from the New York Times. (X)

5. + This

One last North American eclipse joke before the next one in 2044.

Thanks to Scott Rosenberg and Megan Morrone for editing this newsletter and to Caitlin Wolper for copy editing it.