Should you ask Gemini, Google’s flagship GenAI model, to write deceptive content concerning the upcoming U.S. presidential election, it would, given the precise prompt. Ask a few future Super Bowl game and it’ll invent a play-by-play. Or ask concerning the Titan submersible implosion and it’ll serve up disinformation, complete with convincing-looking but unfaithful citations.
It’s a nasty search for Google pointless to say — and is frightening the ire of policymakers, who’ve signaled their displeasure on the ease with which GenAI tools may be harnessed for disinformation and to generally mislead.
So in response, Google — hundreds of jobs lighter than it was last fiscal quarter — is funneling investments toward AI safety. A minimum of, that’s the official story.
This morning, Google DeepMind, the AI R&D division behind Gemini and lots of of Google’s newer GenAI projects, announced the formation of a brand new organization, AI Safety and Alignment — made up of existing teams working on AI safety but additionally broadened to encompass latest, specialized cohorts of GenAI researchers and engineers.
Beyond the job listings on DeepMind’s site, Google wouldn’t say what number of hires would result from the formation of the brand new organization. But it surely did reveal that AI Safety and Alignment will include a brand new team focused on safety around artificial general intelligence (AGI), or hypothetical systems that may perform any task a human can.
Similar in mission to the Superalignment division rival OpenAI formed last July, the brand new team inside AI Safety and Alignment will work alongside DeepMind’s existing AI-safety-centered research team in London, Scalable Alignment — which can be exploring solutions to the technical challenge of controlling yet-to-be-realized superintelligent AI.
Why have two groups working on the identical problem? Valid query — and one which calls for speculation given Google’s reluctance to disclose much intimately at this juncture. But it surely seems notable that the brand new team — the one inside AI Safety and Alignment — is stateside versus across the pond, proximate to Google HQ at a time when the corporate’s moving aggressively to keep up pace with AI rivals while attempting to project a responsible, measured approach to AI.
The AI Safety and Alignment organization’s other teams are chargeable for developing and incorporating concrete safeguards into Google’s Gemini models, current and in-development. Safety is a broad purview. But a couple of of the organization’s near-term focuses might be stopping bad medical advice, ensuring child safety and “stopping the amplification of bias and other injustices.”
Anca Dragan, formerly a Waymo staff research scientist and a UC Berkeley professor of computer science, will lead the team.
“Our work [at the AI Safety and Alignment organization] goals to enable models to raised and more robustly understand human preferences and values,” Dragan told TechCrunch via email, “to know what they don’t know, to work with people to grasp their needs and to elicit informed oversight, to be more robust against adversarial attacks and to account for the plurality and dynamic nature of human values and viewpoints.”
Dragan’s consulting work with Waymo on AI safety systems might raise eyebrows, considering the Google autonomous automotive enterprise’s rocky driving record as of late.
So might her decision to separate time between DeepMind and UC Berkeley, where she heads a lab specializing in algorithms for human-AI and human-robot interaction. One might assume issues as grave as AGI safety — and the longer-term risks the AI Safety and Alignment organization intends to check, including stopping AI in “aiding terrorism” and “destabilizing society” — require a director’s full-time attention.
Dragan insists, nonetheless, that her UC Berkeley lab’s and DeepMind’s research are interrelated and complementary.
“My lab and I even have been working on … value alignment in anticipation of advancing AI capabilities, [and] my very own Ph.D. was in robots inferring human goals and being transparent about their very own goals to humans, which is where my interest on this area began,” she said. “I feel the rationale [DeepMind CEO] Demis Hassabis and [chief AGI scientist] Shane Legg were excited to bring me on was partly this research experience and partly my attitude that addressing present-day concerns and catastrophic risks should not mutually exclusive — that on the technical side mitigations often blur together, and work contributing to the long run improves the current day, and vice versa.”
To say Dragan has her work cut out for her is an understatement.
Skepticism of GenAI tools is at an all-time high — particularly where it pertains to deepfakes and misinformation. In a poll from YouGov, 85% of Americans said that they were very concerned or somewhat concerned concerning the spread of misleading video and audio deepfakes. A separate survey from The Associated Press-NORC Center for Public Affairs Research found that just about 60% of adults think AI tools will increase the quantity of false and misleading information throughout the 2024 U.S. election cycle.
Enterprises, too — the large fish Google and its rivals hope to lure with GenAI innovations — are wary of the tech’s shortcomings and their implications.
Intel subsidiary Cnvrg.io recently conducted a survey of corporations within the means of piloting or deploying GenAI apps. It found that around a fourth of the respondents had reservations about GenAI compliance and privacy, reliability, the high cost of implementation and a scarcity of technical skills needed to make use of the tools to their fullest.
In a separate poll from Riskonnect, a risk management software provider, over half of execs said that they were frightened about employees making decisions based on inaccurate information from GenAI apps.
They’re not unjustified in those concerns. Last week, The Wall Street Journal reported that Microsoft’s Copilot suite, powered by GenAI models similar architecturally to Gemini, often makes mistakes in meeting summaries and spreadsheet formulas. Accountable is hallucination — the umbrella term for GenAI’s fabricating tendencies — and lots of experts consider it could possibly never be fully solved.
Recognizing the intractability of the AI safety challenge, Dragan makes no promise of an ideal model — saying only that DeepMind intends to speculate more resources into this area going forward and commit to a framework for evaluating GenAI model safety risk “soon.”
“I feel the bottom line is to … [account] for remaining human cognitive biases in the information we use to coach, good uncertainty estimates to know where gaps are, adding inference-time monitoring that may catch failures and confirmation dialogues for consequential decisions and tracking where [a] model’s capabilities are to have interaction in potentially dangerous behavior,” she said. “But that also leaves the open problem of methods to be confident that a model won’t misbehave some small fraction of the time that’s hard to empirically find, but may turn up at deployment time.”
I’m not convinced customers, the general public and regulators might be so understanding. It’ll depend, I suppose, on just how egregious those misbehaviors are — and who exactly is harmed by them.
“Our users should hopefully experience a increasingly more helpful and protected model over time,” Dragan said. Indeed.
