Jun 6, 2024 - Axios Events

AI+ NY Summit 2024 Roundup

Axios AI+ NY Summit on June 5, 2024.

Axios AI+ NY Summit 2024. Photo credit: DP Jolly on behalf of Axios.

NEW YORK – Safety concerns and AI's endless potential – from sports analytics to fragrance – were highlights from top tech leaders and entrepreneurs at Axios' AI+ NY Summit on June 5th.

Encrypted messaging app Signal president Meredith Whittaker said that because the cost to create and run AI is high, tech companies will most likely sell people's personal information in order to try to pay for it.

  • "It costs hundreds of millions of dollars to train these models…So there is deep pressure from companies — that are basically promising God and delivering email prompts — to make some return on investment in this technology."

Whittaker also called Microsoft's new Recall feature for PCs "spyware" saying it is "a serious hijacking of trust" and "a dangerous honeypot for hackers."

  • How it works: Microsoft's Recall feature keeps everything someone does on their computer in an AI-searchable database.
  • Many people use Signal on their desktop and expect privacy and encryption, but Whittaker says Microsoft is essentially "snapshotting" the messages.
  • What they're saying: "I think we should be mortified. Especially those of us who have a bit of understanding of what this tech actually does and the track record of these companies and the political environment in which they're operating in."

In another interview, Allen Institute for AI CEO Ali Farhadi spoke about how he thinks the tech industry broke the trust of the public by launching AI while it is still unreliable and lacks proper evaluation.

  • "We are in an evaluation crisis mainly because of the capability of these models. The rate at, or the pace of these capabilities, or us discovering those capabilities severely outpaced our ability to evaluate."

Separately, AI expert Helen Toner told Axios' Ina Fried that people need to think more "holistically" about the negative outcomes that could occur with AI.

  • She explained that sometimes people think of far out scenarios, like "robots shooting humans."
  • "There's a lot of other ways that AI could really take things in a bad direction, and we do ourselves a disservice by locking in these very specific stories."

Safety isn't the only concern industry leaders have, as more creators, including those who earn their livelihoods through online ventures, feel threatened by the AI.

Cookbook author and YouTuber Lisa Bryan and ad management platform Raptive CEO Michael Sanchez discussed how generative AI is using creators' work as training data, replacing them and making it more difficult for audiences to discover individual and unique creator content.

  • Google's newly launched AI overview search results leaves creators experiencing a decline in traffic to their websites, Bryan and Sanchez explained.
  • On top of that, some creators are upset because they believe companies like Google that are using generative AI, are scrapping creators' content to train their models and spit back out an AI answer that's not as good or just wrong. Bryan said it is, "one of the biggest heists of copyright infringement is happening right now."

As a solution, Sanchez wants lawmakers to do something about it.

  • "We've been spending some time talking to the legislators and staff on both sides of the aisle, typically the Senate and the Judiciary...both sides are very concerned about anti-competitive [behavior], copyright and fair use," said Sanchez.

On the bright side, AI is continuing to reach new heights – or more accurately, smells.

Osmo co-founder and CEO Alex Wiltschko discussed the company's innovative technology to digitize smell with the help of AI.

  • What they're saying: "Vision and hearing have already been digitized, but not evolution's first invention: smell. That is our first and most primary and most emotional sense and computers have no idea how to do this."
  • Wiltschko thinks sending a scent or a smell to someone via email or online chat is "going to happen" someday.

Amazon Web Services and the National Football League showcased their Digital Athlete product, which is an AI powered tool to see which players have the potential to be injured, or to even simulate potential rules changes in the league.

  • How it works: The technology is built into players' protective gear and it'll analyze millions of data points as players move around. Using that data, they'll be able to identify for example when players could be at an increased risk of injury.
  • Data is also being taken from cameras that are recording the football games, said Jeff Miller, executive vice president for player safety, communications, public affairs and policy at the NFL.
  • Miller added that clubs, trainers, and doctors also have access to this data. "We can predict their risk profiles. At what point is somebody running hot? Which receiver do you need to pull back on that week? Who's been exposed to more than they should? And we can start to govern how we manage them."

Content from sponsored segment below:

In a View From the Top conversation, IBM senior vice president of marketing and communications Jonathan Adashek and tennis star Maria Sharapova discussed how IBM has been using generative AI with their watsonx technology to collect millions of data points during professional tennis matches.

  • "It's all about taking the data and providing a new way to look at it and get better insights from it," Adashek said.
  • Sharapova explained how all that data being gathered within seconds benefits players looking to step up their game or analyze their opponents techniques.

In a separate View From the Top conversation, Meta vice president of product Connor Hayes highlighted expansion of Meta AI internationally and how the uptake has been "encouraging."

  • But Hayes also expressed that one of the challenges Meta is facing is also an industry wide issue of "alignment or controllability of the model."
  • "In a product like Meta AI, we have a set of content standards of topics that we want it to engage on and not engage on, or sort of the voice and tone of the assistant that you want to maintain consistently. And just dealing with keeping the model aligned to the sort of rules and tone that we've given it has been a big challenge."
Go deeper