Latest / Columbus' Morning News / Amazon reportedly plans major cuts!
Transcript
- 0:00Speaker 1: Want to check in with a Mike Tbuski, ABC News
- 0:02Speaker 1: Technology reporter. I'm a Legacy Retirement Group dot com phone line.
- 0:06Speaker 1: It is Tech Tuesday, and yesterday Mike OpenAI, which is
- 0:10Speaker 1: the chat GPT program. They are now looking at data
- 0:15Speaker 1: that something I think we all kind of knew was coming,
- 0:18Speaker 1: that this is sort of affecting public health when it
- 0:21Speaker 1: comes to AI and chat GPT.
- 0:23Speaker 2: What's the story, Yeah, in a pretty big way from
- 0:25Speaker 2: what we understand. So on Monday, OpenAI released our first
- 0:29Speaker 2: ever look really at some of the hard data related
- 0:32Speaker 2: to this phenomenon known as AI psychosis. Now, in any
- 0:36Speaker 2: given week, according to this new data, point zero seven
- 0:39Speaker 2: percent of chat GPT users experience potential signs of psychosis
- 0:44Speaker 2: or mania. According to the company, point one five percent
- 0:48Speaker 2: of users have conversations that include quote explicit indicators of
- 0:52Speaker 2: potential suicidal planning or intent, and point one five percent
- 0:56Speaker 2: again showed potential heightened levels of emotional attachment to chat GPT.
- 1:01Speaker 2: Of course, this comes against the backdrop of a whole
- 1:04Speaker 2: variety of stories that we've seen reported in the last
- 1:07Speaker 2: few months of AI psychosis affecting people across the country
- 1:11Speaker 2: and around the world. These are stories of people experiencing
- 1:14Speaker 2: severe paranoia, delusions, other mental health crises while using generative AI,
- 1:20Speaker 2: not always just Chat GPT, but Google, Gemini and others.
- 1:24Speaker 2: Usually this is sort of extended periods that they are
- 1:27Speaker 2: spending with these chatbots. And usually these people do have
- 1:30Speaker 2: a history of mental health issues or trauma in their
- 1:34Speaker 2: past that could also have played a role here. But
- 1:37Speaker 2: some of these incidents have taken pretty dark turns, guys.
- 1:40Speaker 2: They've led to people cutting off communication with their friends
- 1:44Speaker 2: and their families, to disappearances in some cases, or even
- 1:48Speaker 2: in extreme cases, people harming themselves and others.
- 1:51Speaker 1: Yeah, that is sobering when you hear that, And I
- 1:54Speaker 1: mean it sounds like though that there are some people
- 1:56Speaker 1: with some predispositions for some mental health problem still though,
- 2:01Speaker 1: and these are people that are are they using AI
- 2:04Speaker 1: for companionship or are they doing it? Are they using
- 2:08Speaker 1: AI for a job or for some other practical purpose.
- 2:12Speaker 2: So again, the reporting here is maybe separate from the data.
- 2:17Speaker 2: Open a Eye doesn't get into specifically what conversations these
- 2:20Speaker 2: people are having or kind of the specifics of any
- 2:23Speaker 2: given case. They do say that they evaluated about eighteen
- 2:26Speaker 2: hundred conversations that touch these sort of dangerous topics. They say,
- 2:31Speaker 2: either thoughts of suicide or you know, drug abuse, you know,
- 2:35Speaker 2: cutting off communication with friends and family, that type of thing.
- 2:39Speaker 2: In the cases that have been reported, which are sort
- 2:42Speaker 2: of more singular, things vary, right. You know, these people
- 2:44Speaker 2: do have you know, seemingly histories of you know, mental
- 2:48Speaker 2: health problems, and they do spend a lot of time
- 2:52Speaker 2: with these chatbots. And the thing with these chatbots is
- 2:55Speaker 2: that they are kind of designed to keep you on
- 2:58Speaker 2: the platform, Right. They're designed to be really friendly and
- 3:00Speaker 2: really nice and try to be really helpful. But over
- 3:03Speaker 2: time they start to break down. Right, They're not really
- 3:06Speaker 2: meant to be used for hours and hours at a time.
- 3:08Speaker 2: They're meant to be used for a quick couple of questions,
- 3:12Speaker 2: maybe to help you understand a project or whatever it
- 3:14Speaker 2: happens to be. Over time, they start to lose their accuracy,
- 3:18Speaker 2: which has always been a real bugbear for this industry.
- 3:21Speaker 2: And you know, given some of this new data on
- 3:24Speaker 2: mental health that we're seeing, it really does run this
- 3:26Speaker 2: pretty substantial risk, especially to the general public.
- 3:29Speaker 1: Any discussion of putting some guardrails in place, maybe some
- 3:33Speaker 1: time limits of hey, you've been talking for you know,
- 3:36Speaker 1: half an hour, forty five minutes, it's time to go
- 3:39Speaker 1: out and look at the blue sky and see the
- 3:41Speaker 1: real world for a minute, go touch grass.
- 3:42Speaker 2: In other words, exactly so to a degree, open Aye
- 3:46Speaker 2: already does this. It's worth mentioning. They, you know, do
- 3:49Speaker 2: have certain signals and flags that go up when they
- 3:52Speaker 2: sense that someone is maybe touching one of these dangerous topics,
- 3:56Speaker 2: and they will point the user to mental health resources.
- 4:00Speaker 2: Of course, that's not a great system, it's not a
- 4:02Speaker 2: perfect system, because these AI systems do hallucinate, they go wrong,
- 4:06Speaker 2: and as we said, exposed use to them, you know,
- 4:08Speaker 2: extended use to these AI chatbots, you know, that is
- 4:12Speaker 2: going to become less and less likely. So that's one
- 4:15Speaker 2: piece of this. The other piece of this is that
- 4:17Speaker 2: open Ai does say their latest technology improves on these
- 4:20Speaker 2: negative outcomes, these experiences where chat GPT will kind of,
- 4:25Speaker 2: you know, support wrongheaded thinking, thinking that goes awry or
- 4:29Speaker 2: leads you to some dark places. They say GPT five,
- 4:32Speaker 2: which is sort of the core technology that underpins chat GPT,
- 4:36Speaker 2: they reduce undesired answers between forty and fifty percent when
- 4:40Speaker 2: compared to the previous model. They've also cross referenced all
- 4:43Speaker 2: this with about one hundred and seventy scientists, primary care physicians,
- 4:47Speaker 2: and psychologists who basically say they don't always agree with
- 4:50Speaker 2: what chat GPT is saying in terms of mental health,
- 4:52Speaker 2: but that it is an improvement over what it was
- 4:55Speaker 2: just a few years ago.
- 4:56Speaker 1: Mike Dubuski, ABC News Technology Reporter, It's funny, MICUs I
- 4:59Speaker 1: was talking to somebody over the weekend about this very thing.
- 5:02Speaker 1: It seems like no one loves AI. That they but
- 5:08Speaker 1: it's like almost necessary evil that no one loves the
- 5:11Speaker 1: prospect of relying on this for the next you know, five, ten, fifteen,
- 5:16Speaker 1: twenty fifty years. But on the other hand, he talked
- 5:20Speaker 1: to these guys, you know, professionally, they all use it
- 5:23Speaker 1: every day. It's become a part of our systems in
- 5:27Speaker 1: whatever it is you do for a living that you can,
- 5:29Speaker 1: you know, refine a speech or a presentation. No one
- 5:32Speaker 1: loves that they have to do that, but it's so easy,
- 5:34Speaker 1: and it's so and mostly very effective.
- 5:38Speaker 2: Yeah, it's a really interesting debate that's playing out right now,
- 5:41Speaker 2: and I think a lot of that also comes back to,
- 5:43Speaker 2: you know, whether companies mandate the use of AI or
- 5:46Speaker 2: integrate AI into their workforces and how that affects how
- 5:50Speaker 2: people use it. I do think that there's a conversation
- 5:54Speaker 2: worth being had here about whether AI is actually going
- 5:57Speaker 2: to be the future. You know, again, we've talked in
- 5:59Speaker 2: the past about this concept of an AI bubble. There's
- 6:02Speaker 2: a lot of money floating around this space and not
- 6:04Speaker 2: a ton of profit. Open Ai, for example, is a
- 6:06Speaker 2: deeply unprofitable company, despite the fact that we talk about
- 6:09Speaker 2: them all the time and they are kind of at
- 6:11Speaker 2: the forefront of this generative AI game right now. Eventually
- 6:15Speaker 2: the bills are going to come to, right like, eventually
- 6:17Speaker 2: they're going to have to start paying back Oracle and
- 6:20Speaker 2: and Nvidia and all these other companies, and you know
- 6:23Speaker 2: that that date is going to come within the next
- 6:25Speaker 2: few years. And if they don't have a product that
- 6:27Speaker 2: people are going to reliably spend money on, well then
- 6:30Speaker 2: that's going to be a pretty you know, the you know,
- 6:32Speaker 2: purse string tightening moment for this this major company. And
- 6:36Speaker 2: I think there's some real questions about whether they can
- 6:38Speaker 2: actually pay the bills once they do come to So
- 6:40Speaker 2: I think that, yeah, sure, it is an interesting debate
- 6:43Speaker 2: as to whether people are going to use it or not.
- 6:44Speaker 2: I think it's also an interesting debate as to whether
- 6:46Speaker 2: we're going to be talking about AI in three or
- 6:48Speaker 2: four years.
- 6:49Speaker 1: Yeah, there could be this major backlash and or I
- 6:52Speaker 1: will see something like maybe ad supported AI, which they're
- 6:55Speaker 1: already talking about. Is it really, oh wow, unbelievable. Before
- 6:58Speaker 1: I let you go, what's going on with Amazon? They're
- 7:00Speaker 1: reportedly preparing to make some staffing cuts I think today.
- 7:04Speaker 2: Yeah, in fact, they've already announced it, about fourteen thousand jobs.
- 7:07Speaker 2: Amazon says they're going to cut as they work on quote,
- 7:10Speaker 2: reducing bureaucracy, removing layers, and shifting resources. That's according to
- 7:15Speaker 2: a statement out of Amazon. Interestingly enough, this is actually
- 7:17Speaker 2: a slightly smaller reduction in workforce than what we were expecting.
- 7:20Speaker 2: Reuter's was the first to report this yesterday when they
- 7:23Speaker 2: said that they were expecting cuts around thirty thousand Amazon workers.
- 7:27Speaker 2: Even still, fourteen thousand people accounts for about four percent
- 7:30Speaker 2: of the company. Writ large, about three hundred and fifty
- 7:33Speaker 2: thousand people work for Amazon. This also comes against the
- 7:36Speaker 2: backdrop of Andy Jassey, the CEO, saying that you know,
- 7:40Speaker 2: he wants to, you know, make this company behave a
- 7:42Speaker 2: little bit more like an AI startup replace some jobs
- 7:45Speaker 2: with artificial intelligence, and the World Economic Times recently found
- 7:49Speaker 2: that forty percent of employers plan to reduce staff and
- 7:52Speaker 2: replace them with AI. But again, given some of the
- 7:54Speaker 2: problems that we've been discussing this morning, you know, hallucination,
- 7:57Speaker 2: not the least of which you know, there's some risk
- 7:59Speaker 2: that comes along with that as well.