Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
ABC News Technology reporter Mike Dubuski is joining us now,
and uh does that?
Speaker 2 (00:06):
Does that raise an eyebrow? Mike? When you hear chili
flavored ice cream? Are you like, uh? No, thank you?
Speaker 3 (00:12):
I will say, guys, the main thought that I had
was I'm glad that you're where you are and I
am doing this remotely.
Speaker 4 (00:19):
The after effects are frightening, aren't they. I mean the
possibilities here of chili flavored ice cream?
Speaker 3 (00:24):
Yeah, Man, science will never know or maybe they will
later to that.
Speaker 1 (00:29):
Yeah, if you were here in studio with us doing
a hit here in you know, live, I'd be like, hey, Mike,
why don't you join us to try this?
Speaker 2 (00:36):
And you go, you do it. You probably would do
what Chuck just did, which is, let me check my contract. No,
I won't be.
Speaker 1 (00:42):
Doing it, so yeah, and I'm excited to do it.
I love stuff like this because it's it's a new thing.
And you know, Skyline Chili is this area of the
country thing.
Speaker 2 (00:55):
I don't know if you've ever had it, heard of it.
Speaker 3 (00:58):
I've not had of it, but had it but lived
in I'm aware of it.
Speaker 1 (01:01):
Okay, very cool, and but then you combine it with
ice cream, and you go, what tarnation is going on
with this?
Speaker 2 (01:08):
You know?
Speaker 1 (01:08):
So it's yeah, well we'll see if if it's a
marriage made in heaven or what have you.
Speaker 2 (01:14):
It's kind of interesting.
Speaker 4 (01:15):
You have any tarnation flavored ice cream? I try some
of thatation.
Speaker 1 (01:20):
Yeah, I don't know, but Greaters is big in this
area too, so we'll see.
Speaker 2 (01:26):
I have a sneak in suspicion.
Speaker 1 (01:27):
It's probably gonna be sneaky good because in their chili
is you know that it's distinct. It's like a cinnamon
kind of finish to the chili, and it's very thin
and all of those things.
Speaker 2 (01:38):
But man, is it really good? And I can't.
Speaker 1 (01:42):
Eat that all the time. Their dogs they're chili dogs
or what have you, because they tear me up. But
I do like the taste of them. They're They're fantastic,
no question about that. So all right, So onto what
we actually had you coming on to kind of unpack
for us here.
Speaker 4 (01:57):
Mike China, What would Chinese Ais say about chili flavored
ice cream? That's what we want, you know.
Speaker 3 (02:02):
I haven't gotten a chance to ask it yet.
Speaker 1 (02:05):
Yeah, that's a chat GPT thing, right, like, yeah, yeah,
see what what would what would come back but talk
to us about this Deep Seek because I think this
is raising some eyebrows.
Speaker 3 (02:16):
Isn't it? It very much is really across a number
of different sectors, both Silicon Valley, Washington, DC, and Wall Street.
So deep Seek is a Chinese research firm that has
come out with an artificial intelligence model recently that it
says can compete with the best American AI models for
just a fraction of the cost. So this model is
(02:39):
called R one. It's a reasoning model and on popular
AI tests so logic puzzles and that sort of thing,
this either matches or beats the scores that open ai
and Meta and Amazon and others have been able to manage.
So that would be one thing. If that were it,
that would be enough to worry a lot of people
(03:00):
in Pillicon Valley. But the more important piece here is
that deep Seek says they were able to train this
model for just six million dollars. Now, for comparison, guys,
Meta spent about ten times that amount on their equivalent model,
about sixty five million dollars, and open ai recently said
that they spent more than one hundred million dollars developing
(03:21):
GPT four, which is the name for their model. Now
we're taking Deep Seek at their word on a lot
of this, which is important, but even still it has
clearly spooked a lot of investors.
Speaker 1 (03:31):
Man, Yeah, this is uh, this is definitely the fact
that it's cheaper.
Speaker 2 (03:39):
I mean, this is I don't know, gosh, I don't
even know.
Speaker 4 (03:44):
The name scares me. I'm gonna be honest with you all.
That's what you were saying, Chinese software. It's going to
be popular because it's cheaper than American versions, and it's
called deep Seek. Putting that in my machine scares me.
I'm just being honest. That name is frightening.
Speaker 3 (04:02):
Well, I mean, naming convention aside. I think there is
really room here for some skepticism and for some caution, right.
You know, in the past, we've seen Chinese tech companies
kind of be used as a means by which to spy,
you know. Of course, that was a lot of the
conversation around TikTok not so long ago. And what's more,
there's reason to be skeptical just on the merits of
(04:24):
what Deep Seek is claiming here. In particular, many people
that I've talked to are especially raising their eyebrows around
that six million dollar figure. They want to know where
that came from. But if we zoom out here just
a little bit, guys, the Biden administration a few years
ago severely limited chip manufacturers. These are the computer processor
makers like Nvidia from sending out their most powerful AI
(04:46):
chips to China. The effort here was to reduce China's
ability to develop AI that would supersede hours. However, if
deep Seak is to be believed here, that strategy backfire.
Chinese engineers committed to using cheap software, readily available software,
and other methods to basically get really good at making
efficient technology. Hence a model that is equivalent to the
(05:10):
best America has to offer using just one one twentieth
of the processing power. We think, so that is part
of this. However, others have speculated that they could be
relying on newer chips that somehow made their way into
China that were smuggled there, which is something that we
know happens. So a lot of questions left to be
answered around this, So I think it's important to treat
(05:32):
it with a big old shaker of salt.
Speaker 1 (05:33):
I was going to say, is the Wall Street being
worried about this. Is that legit based on this or
is it something that Well, they get worried about a
lot of things. But then some people go, hey, it's
legit when Wall Street is worrying about something. You know,
But I don't know, I don't know where you come
down on this.
Speaker 3 (05:53):
I mean, if I knew the answer to that question,
I think my wallet would be certainly heavier this after.
But yeah, I mean, you're absolutely right to say that
Wall Street does get a little nervous sometimes when it
doesn't necessarily need to. For example, some tech stocks actually
aren't down today. Meta for example, is slightly up despite
the fact that this deep Seek model by you know,
(06:14):
common testing is completely outpaced, or rather deep Seeks model
outpaces Meta's model. And the thinking there is that, you know,
the arising tide lifts all boats, right, Competition is good
for the industry. The fact that China was able to
make this advancement, if it ends up being true, is
going to make it possible and maybe light a fire
(06:35):
under these other companies. So that is, you know, potentially
cause for optimism in the II space. On the other hand,
it could pop the bubble that many think is happening
in the II space right now. So that is kind
of where investors come down on this. But the big
takeaway here, guys, I think is important to remember last week,
the prevailing wisdom in the AI space was that more
(06:56):
is better, right, more training data, more computing power, more
more energy. That's ultimately going to give you an AI
technology that's going to radically change the world. Right. What
this Deep Seek model seems to be saying is that
less is more. Right. They can do the same or
better than American models with a lot less and that
(07:19):
opens the industry up to a lot of new players.
After all, there are a lot of companies with access
to six million dollars. There are a lot fewer companies
with access to one hundred million dollars.
Speaker 1 (07:29):
So as far as the timing, then we it just
remains to be seen. As far as if this really
is something that you know, is is it?
Speaker 2 (07:37):
You know?
Speaker 1 (07:38):
Danger danger right now that Wall Street is it's it's
a timing thing and there's really no way to tell
when that will happen.
Speaker 3 (07:45):
Yeah, it's a waiting game at this point. I mean,
at this point, the app is very popular. The app
version of this model is topping app store charts, so
people seem to be interested in it to a degree
that Deep Seak is actually pausing new signups. They say
that they've been hit with a side attack because there's
too much attention on them right now. So it's an
evolving situation. We're going to track it throughout the week,
of course.
Speaker 2 (08:06):
Very interesting.
Speaker 1 (08:06):
ABC News Technology reporter Mike Dubuski, Mike, thank you very much.
Speaker 2 (08:10):
Appreciate you man.
Speaker 4 (08:11):
Of course, See, you Americans need to get better at that.
Asian countries have been doing it to us for years.
I mean since the invention of the transistor. They have
basically taken what we do, duplicate it much cheaper, and
then beat us over the head with our own stick.
And it wasn't that long ago we didn't want video
games being shipped some places because we didn't want the
(08:33):
chips in video games being made available to people who
were developing this kind of thing. And here comes to
bite us again.
Speaker 2 (08:39):
Yeah.
Speaker 1 (08:40):
So then based on the way you kind of just
surmised everything, it's like Wall Street worrying is legit.
Speaker 4 (08:47):
Yeah, unless everybody on Wall Street starts buying into Chinese
companies that are beating us with our own stick, yeah,
I guess.