Dr. Geoffrey Hinton, who won the Turing Award nearly a decade ago and is now a former Google employee, is known for shouting the mania among AI students. But now that the “Godfather of AI” has moved away from Google, we’re hearing more and more desperate AI alarm bells.
In an exit interview with
new york times (opens in new tab) , Hinton expressed deep concern about the rapid expansion of AI, saying, “It’s hard to understand how we can prevent bad actors from exploiting AI.”
Hinton’s decades of pioneering work in neural networks and today’s chatbots (ChatGPT, Google Bard, Bing AI) are a direct line. His breakthrough over a decade ago led Google to involve him in the development of the next generation of deep learning systems. The system helps computers interpret images, text, and speech in the same way humans do.
talk to W
annoyed in 2014 (opens in new tab), Hinton’s enthusiasm was evident. “I’m very excited when we discover ways to make neural networks better, and when it goes hand-in-hand with how the brain works.”
A very different Hinton spoke
new york times (opens in new tab) This week we outline all the ways AI can keep humanity on track. The main points are: Rush to Competition Means Guardrails Splash
Companies such as Microsoft, Google and OpenAI often profess to take a slow and cautious approach to AI development and chatbot deployment, Dr. Hinton said.
new york times He said he was concerned about the reality that increasing competition is undermining a cautious approach. I’ve seen Google rush out a bard not ready for primetime to respond.
Will these companies be able to balance the market imperative to stay ahead of the competition (for example, Google staying number one in search) and greater profits? Dr. Hinton now agrees. I have not.
loss of truth
Dr. Hinton worries that the proliferation of AI will lead to a deluge of fake online content. Of course, it’s less of a future concern than a real-time concern. Because now people are deceived by AI music, AI news images forging vocal gifts of masters (including dead ones).
(opens in new tab) Treated as authentic, generative image wins photo contest (opens in new tab)With the power and prevalence of deep fakes, the videos we watch today are rarely taken at face value.
Still, Dr. Hinton may be right that this is just the beginning. You can’t trust what you see or hear unless Google, Microsoft, OpenAI, and other companies do something about it.
devastated job market
Dr. Hinton warned
Times That AI is set up to take on more than just the tasks we don’t want to do.
Many of us use ChatBots like Bard and ChatGPT to write presentations, proposals, and even programming. Most of the outputs aren’t prime time ready, but some are, or at least passable.
With dozens of AI-generated novels now on sale on Amazon, the Writers Guild of America has raised concerns studios could outsource work to AI if they don’t agree to the new deal. is expressed. And while there hasn’t been widespread job cuts directly related to AI, the growth of these powerful tools is causing some to rethink their workforce.
Unexpected and unwanted behavior
One of the characteristics of neural nets and deep learning AI is the ability to learn using huge amounts of data. One of the unintended consequences of this human brain-like power is that AI can learn lessons it never expected. A self-determining AI might run AI on those lessons. Dr. Hinton said he is particularly concerned about AI that can do its own programming, not just program it, as it is associated with unintended consequences.
AI will be smarter than us
Today’s AIs often appear smarter than humans, but their penchant for hallucinations and fabricated facts leaves them far from our great brains. Dr. Hinton believes the day when AI can outsmart us is fast approaching.
AI could fake empathy – see recent study as source of medical advice
(opens in new tab) – But that’s not the same as true human empathy. A super smart system that knows everything but doesn’t consider how its choices affect humans is very concerning.
Dr. Hinton’s caveats are in stark contrast to his initial enthusiasm for the tech sector he invented, but he says:
Wired A 2014 study of an AI system that mimics the human brain sounds strangely prescient now.