AI

Nvidia’s Jensen Huang says AI hallucinations are solvable, artificial general intelligence is 5 years away

Comment

Image Credits: Haje Jan Kamps (opens in a new window)

Artificial general intelligence (AGI) — often referred to as “strong AI,” “full AI,” “human-level AI” or “general intelligent action” — represents a significant future leap in the field of artificial intelligence. Unlike narrow AI, which is tailored for specific tasks, such as detecting product flaws, summarizing the news, or building you a website, AGI will be able to perform a broad spectrum of cognitive tasks at or above human levels. Addressing the press this week at Nvidia’s annual GTC developer conference, CEO Jensen Huang appeared to be getting really bored of discussing the subject — not least because he finds himself misquoted a lot, he says.

The frequency of the question makes sense: The concept raises existential questions about humanity’s role in and control of a future where machines can outthink, outlearn and outperform humans in virtually every domain. The core of this concern lies in the unpredictability of AGI’s decision-making processes and objectives, which might not align with human values or priorities (a concept explored in-depth in science fiction since at least the 1940s). There’s concern that once AGI reaches a certain level of autonomy and capability, it might become impossible to contain or control, leading to scenarios where its actions cannot be predicted or reversed.

When sensationalist press asks for a timeframe, it is often baiting AI professionals into putting a timeline on the end of humanity — or at least the current status quo. Needless to say, AI CEOs aren’t always eager to tackle the subject.

Huang, however, spent some time telling the press what he does think about the topic. Predicting when we will see a passable AGI depends on how you define AGI, Huang argues, and draws a couple of parallels: Even with the complications of time zones, you know when New Year happens and 2025 rolls around. If you’re driving to the San Jose Convention Center (where this year’s GTC conference is being held), you generally know you’ve arrived when you can see the enormous GTC banners. The crucial point is that we can agree on how to measure that you’ve arrived, whether temporally or geospatially, where you were hoping to go.

“If we specified AGI to be something very specific, a set of tests where a software program can do very well — or maybe 8% better than most people — I believe we will get there within 5 years,” Huang explains. He suggests that the tests could be a legal bar exam, logic tests, economic tests or perhaps the ability to pass a pre-med exam. Unless the questioner is able to be very specific about what AGI means in the context of the question, he’s not willing to make a prediction. Fair enough.

AI hallucination is solvable

In Tuesday’s Q&A session, Huang was asked what to do about AI hallucinations — the tendency for some AIs to make up answers that sound plausible but aren’t based in fact. He appeared visibly frustrated by the question, and suggested that hallucinations are solvable easily — by making sure that answers are well-researched.

“Add a rule: For every single answer, you have to look up the answer,” Huang says, referring to this practice as “retrieval-augmented generation,” describing an approach very similar to basic media literacy: Examine the source and the context. Compare the facts contained in the source to known truths, and if the answer is factually inaccurate — even partially — discard the whole source and move on to the next one. “The AI shouldn’t just answer; it should do research first to determine which of the answers are the best.”

For mission-critical answers, such as health advice or similar, Nvidia’s CEO suggests that perhaps checking multiple resources and known sources of truth is the way forward. Of course, this means that the generator that is creating an answer needs to have the option to say, “I don’t know the answer to your question,” or “I can’t get to a consensus on what the right answer to this question is,” or even something like “Hey, the Super Bowl hasn’t happened yet, so I don’t know who won.”

Catch up on Nvidia’s GTC 2024:

More TechCrunch

Google on Thursday said it is rolling out NotebookLM, its AI-powered note-taking assistant, to over 200 new countries, nearly six months after opening its access in the U.S. The platform,…

Google’s updated AI-powered NotebookLM expands to India, UK and over 200 other countries

Inflation and currency devaluation have always been a growing concern for Africans with bank accounts.

Once serving war-torn Sudan, YC-backed Elevate now provides fintech to freelancers globally

Featured Article

Amazon buys Indian video streaming service MX Player

Amazon has agreed to acquire assets of Indian video streaming service MX Player from the local media powerhouse Times Internet, the latest step by the e-commerce giant to make its services and brand popular in smaller cities and towns in the key overseas market.  The two firms reached a definitive…

3 hours ago
Amazon buys Indian video streaming service MX Player

Dealt is now building a service platform for retailers instead of end customers.

Dealt turns retailers into service providers and proves that pivots sometimes work

Snowflake is the latest company in a string of high-profile security incidents and sizable data breaches caused by the lack of MFA.

Hundreds of Snowflake customer passwords found online are linked to info-stealing malware

The buy will benefit ChromeOS, Google’s lightweight Linux-based operating system, by giving ChromeOS users greater access to Windows apps “without the hassle of complex installations or updates.”

Google acquires Cameyo to bring Windows apps to ChromeOS

Mistral is no doubt looking to grow revenue as it faces considerable — and growing — competition in the generative AI space.

Mistral launches new services and SDK to let customers fine-tune its models

The warning for the Ai Pin was issued “out of an abundance of caution,” according to Humane.

Humane urges customers to stop using charging case, citing battery fire concerns

The keynote will be focused on Apple’s software offerings and the developers that power them, including the latest versions of iOS, iPadOS, macOS, tvOS, visionOS and watchOS.

Watch Apple kick off WWDC 2024 right here

As WWDC 2024 nears, all sorts of rumors and leaks have emerged about what iOS 18 and its AI-powered apps and features have in store.

What to expect from Apple’s AI-powered iOS 18 at WWDC 2024

Welcome to Elon Musk’s X. The social network formerly known as Twitter where the rules are made up and the check marks don’t matter. Or do they? The Tesla and…

Elon Musk’s X: A complete timeline of what Twitter has become

TechCrunch has kept readers informed regarding Fearless Fund’s courtroom battle to provide business grants to Black women. Today, we are happy to announce that Fearless Fund CEO and co-founder Arian…

Fearless Fund’s Arian Simone coming to Disrupt 2024

Bridgy Fed is one of the efforts aimed at connecting the fediverse with the web, Bluesky and, perhaps later, other networks like Nostr.

Bluesky and Mastodon users can now talk to each other with Bridgy Fed

Zoox, Amazon’s self-driving unit, is bringing its autonomous vehicles to more cities.  The self-driving technology company announced Wednesday plans to begin testing in Austin and Miami this summer. The two…

Zoox to test self-driving cars in Austin and Miami 

Called Stable Audio Open, the generative model takes a text description and outputs a recording up to 47 seconds in length.

Stability AI releases a sound generator

It’s not just instant-delivery startups that are struggling. Oda, the Norway-based online supermarket delivery startup, has confirmed layoffs of 150 jobs as it drastically scales back its expansion ambitions to…

SoftBank-backed grocery startup Oda lays off 150, resets focus on Norway and Sweden

Newsletter platform Substack is introducing the ability for writers to send videos to their subscribers via Chat, its private community feature, the company announced on Wednesday. The rollout of video…

Substack brings video to its Chat feature

Hiya, folks, and welcome to TechCrunch’s inaugural AI newsletter. It’s truly a thrill to type those words — this one’s been long in the making, and we’re excited to finally…

This Week in AI: Ex-OpenAI staff call for safety and transparency

Ms. Rachel isn’t a household name, but if you spend a lot of time with toddlers, she might as well be a rockstar. She’s like Steve from Blues Clues for…

Cameo fumbles on Ms. Rachel fundraiser as fans receive credits instead of videos  

Cartwheel helps animators go from zero to basic movement, so creating a scene or character with elementary motions like taking a step, swatting a fly or sitting down is easier.

Cartwheel generates 3D animations from scratch to power up creators

The new tool, which is set to arrive in Wix’s app builder tool this week, guides users through a chatbot-like interface to understand the goals, intent and aesthetic of their…

Wix’s new tool taps AI to generate smartphone apps

ClickUp Knowledge Management combines a new wiki-like editor and with a new AI system that can also bring in data from Google Drive, Dropbox, Confluence, Figma and other sources.

ClickUp wants to take on Notion and Confluence with its new AI-based Knowledge Base

New York City, home to over 60,000 gig delivery workers, has been cracking down on cheap, uncertified e-bikes that have resulted in battery fires across the city.  Some e-bike providers…

Whizz wants to own the delivery e-bike subscription space, starting with NYC

This is the last major step before Starliner can be certified as an operational crew system, and the first Starliner mission is expected to launch in 2025. 

Boeing’s Starliner astronaut capsule is en route to the ISS 

TechCrunch Disrupt 2024 in San Francisco is the must-attend event for startup founders aiming to make their mark in the tech world. This year, founders have three exciting ways to…

Three ways founders can shine at TechCrunch Disrupt 2024

Google’s newest startup program, announced on Wednesday, aims to bring AI technology to the public sector. The newly launched “Google for Startups AI Academy: American Infrastructure” will offer participants hands-on…

Google’s new startup program focuses on bringing AI to public infrastructure

eBay’s newest AI feature allows sellers to replace image backgrounds with AI-generated backdrops. The tool is now available for iOS users in the U.S., U.K., and Germany. It’ll gradually roll…

eBay debuts AI-powered background tool to enhance product images

If you’re anything like me, you’ve tried every to-do list app and productivity system, only to find yourself giving up sooner rather than later because managing your productivity system becomes…

Hoop uses AI to automatically manage your to-do list

Asana is using its work graph to train LLMs with the goal of creating AI assistants that work alongside human employees in company workflows.

Asana introduces ‘AI teammates’ designed to work alongside human employees

Taloflow, an early stage startup changing the way companies evaluate and select software, has raised $1.3M in a seed round.

Taloflow puts AI to work on software vendor selection to reduce costs and save time