Google I/O unveiled an artificial intelligence assistant that can see and hear the world, while OpenAI unveiled its version she– Just like a chatbot in iPhone. Next week, Microsoft is hosting a Build conference, and there’s bound to be some version of Copilot or Cortana that understands pivot tables. Then, in a few weeks, Apple will hold its own developer conference, which, if lively, will also discuss artificial intelligence. (It’s unclear whether Siri will be mentioned.)
Artificial intelligence is here! It is no longer conceptual. It takes away jobs, creates some new ones, and helps millions of students avoid homework.According to most major tech companies investing in artificial intelligence, we It seems to be at the beginning of one of the rare major technological changes. Think of the Industrial Revolution or the creation of the Internet or the personal computer. All of Silicon Valley — the big tech companies — are focused on taking large language models and other forms of artificial intelligence and moving them from researchers’ laptops to regular people’s phones and computers. Ideally, they’ll make a lot of money in the process.
But I can’t really care about that because Meta AI thinks I have a beard.
I want to be very clear: I am a cisgender woman and do not have a beard. But if I type “show me a picture of Alex Kranz” into the prompt window, Meta AI inevitably returns an image of a very handsome dark-haired man with a beard. I’m just part of it!
Meta AI isn’t the only company struggling with the details edgeheader. ChatGPT told me yesterday that I wasn’t working edge.Google’s Gemini didn’t know who I was (fair enough), but told me Nilay Patel was edge, It then apologized and corrected itself, saying he wasn’t. (I guarantee you he is.)
Artificial intelligence keeps screwing up because these computers are stupid. Their abilities are extraordinary, but their stupidity is astonishing. I couldn’t be more excited about the next tipping point in the artificial intelligence revolution, where computers won’t be able to consistently be accurate about even tiny things.
I mean, they even blew it during Google’s big AI keynote at I/O. In an ad for Google’s new artificial intelligence search engine, someone asked how to fix a stuck film camera, suggesting they “open the back door and gently remove the film.” This is the easiest way to destroy photos you have taken.
The difficult relationship between artificial intelligence and reality is known as “illusion.” In short: These machines are very good at spotting patterns in information, but they occasionally make mistakes when trying to extrapolate and create. They effectively “hallucinate” a new reality, which is often false. This is a tricky problem, and everyone working on artificial intelligence now realizes it.
A former Google researcher claims the issue may be fixed within the next year (although he regrets the outcome), while Microsoft has made a tool available to some of its users that should help detect them.Liz Reid, head of search at Google, said: edge It is also aware of the challenges. She told my colleague David Pierce that any language model “needs to strike a balance between creativity and authenticity.” “We’re really going to tilt it toward the factual side.”
But notice how Reid says there is a balance?That’s because many AI researchers don’t actually think hallucinations are Can solved. A study from the National University of Singapore shows that hallucinations are an inevitable consequence of all large language models. Just as no one is 100 percent right all the time, neither are these computers.
That’s probably why most of the major players in the field—those with the real resources and financial incentive for us all to embrace artificial intelligence—think you shouldn’t worry about it. At Google’s IO keynote, it added the phrase “Check responses for accuracy” in tiny gray font to the screen below nearly every new AI tool it showed off — a helpful reminder that Its tools can’t be trusted, but it doesn’t do that either. ChatGPT operates similarly. The message in small font at the bottom of the prompt window reads: “ChatGPT may have an error. Check important information.”
This is not the disclaimer you want to see on a tool that will change our entire lives in the near future! The people who make these tools don’t seem to care much about fixing the problem other than a small caveat.
OpenAI CEO Sam Altman, who was briefly fired for putting profits over safety, went a step further and said anyone who has questions about the accuracy of artificial intelligence is naive. “If you just do something naive and say, ‘Never say anything you’re not 100 percent sure of,’ you can get them all to do it. But it won’t have the magic that people like so much,” he told an audience at Salesforce’s Dreamforce conference last year.
There is an unquantifiable magic in artificial intelligence that allows us to forgive its tenuous relationship with reality, and the idea is advanced by those eager to escape concerns about accuracy. Google, OpenAI, Microsoft, and many other AI developers and researchers all consider hallucinations a minor annoyance that should be forgiven as they work to create digital beings that make our lives easier.
But with apologies to Sam and everyone else who financially incentivized me to get excited about artificial intelligence. I didn’t get into computers for the inaccurate magic of human consciousness. I came to them because they are very accurate, whereas humans are not. I don’t need my computer to be my friend; I need it to correctly determine my gender when I ask and to help me avoid accidentally exposing my film while repairing a broken camera. I think lawyers want the case law to be correct.
I understand Sam Altman and other AI evangelists are from there. In the distant future, it may be possible to create true digital consciousness from 1 to 0. The current development speed of artificial intelligence is astounding, dwarfing many previous technological revolutions. Now Silicon Valley is working its real magic.
But the artificial intelligence thinks I have a beard. It cannot consistently solve the simplest of tasks, and yet it is thrust upon us, expecting us to celebrate the incredibly mediocre services these artificial intelligences provide. While I certainly marvel at the technological innovations taking place, I wish my computer didn’t sacrifice accuracy so I could have a digital avatar to talk to. It’s not a fair exchange – it’s just a fun exchange.
