She also urged Google to take more responsibility over its AI products and their accuracy
Not really possible when the technology is built around probability.
It is simply *not* deterministic. The results are based on a roll of the dice. They lack consistency.
Ask the same question in a slightly different way or at different time and get a different answer. And not just different wording but also different meaning.
Would you store food in a refrigerator that might *not* keep it from spoiling? Marketing such a product would create a liability issue --- and I expect the same for AI.
The fact that we're investing $ trillions in such a product will not end well.
It's hard not to trust answers which are superhumanly plausible.
Imagine being offered a choice of two oracles. The first oracle would give you an answer that is inspiring yet mysterious and oblique. The second gives an answer that is guaranteed to be unoriginal but also highly, highly plausible. It also maintains a record of all the questions and makes them available to unknown persons.
Should I know, I mistakenly trusted Google when it told me there was a live music event last Sunday in a local bar, turns out Google's AI got confused because there is another bar with the same name in a different city here in Colombia, so I dressed up and everything just to go to a bar that was closed that day.
> However, some experts say big tech firms such as Google should not be inviting users to fact-check their tools' output, but should focus instead on making their systems more reliable.
That is a very strange thing to say. They both are and should be doing both.
She also urged Google to take more responsibility over its AI products and their accuracy
Not really possible when the technology is built around probability.
It is simply *not* deterministic. The results are based on a roll of the dice. They lack consistency.
Ask the same question in a slightly different way or at different time and get a different answer. And not just different wording but also different meaning.
Would you store food in a refrigerator that might *not* keep it from spoiling? Marketing such a product would create a liability issue --- and I expect the same for AI.
The fact that we're investing $ trillions in such a product will not end well.
All generative AI output should begin and end with "Trust this output as much as you would trust a 5-year-old to control your life".
And 4chan has (had? Haven't been there in a while) "only a fool would take the things here as fact"
Doesn't matter, a good chunk of people are going to take everything at face value
It's hard not to trust answers which are superhumanly plausible.
Imagine being offered a choice of two oracles. The first oracle would give you an answer that is inspiring yet mysterious and oblique. The second gives an answer that is guaranteed to be unoriginal but also highly, highly plausible. It also maintains a record of all the questions and makes them available to unknown persons.
I literally have no idea what you mean
> superhumanly plausible
What? Plenty of humans are competent confabulators.
Should I know, I mistakenly trusted Google when it told me there was a live music event last Sunday in a local bar, turns out Google's AI got confused because there is another bar with the same name in a different city here in Colombia, so I dressed up and everything just to go to a bar that was closed that day.
Definitely don't trust the overviews for anything super current. But I suspect you won't anymore :)
> However, some experts say big tech firms such as Google should not be inviting users to fact-check their tools' output, but should focus instead on making their systems more reliable.
That is a very strange thing to say. They both are and should be doing both.
If it's untrustworthy, why do they prevent us from disabling "AI overview"?
Completely at odds with everything the company is actually doing. Shameful nonsense.
If trusting the AI has brought us to this, of what use was the AI?
You can bioil this down to "Don't trust AI."
should we blindly trust what he says
Not sure, let me go ask AI
... too late
and don't blindly believe a bbc headline these days...