1
00:00:07,133 --> 00:00:10,453
Speaker 1: You're listening to the Saturday Morning with Jack team podcast

2
00:00:10,573 --> 00:00:11,693
Speaker 1: from News Talks at be.

3
00:00:12,733 --> 00:00:15,333
Speaker 2: The next level of AI is about to be released

4
00:00:15,333 --> 00:00:18,453
Speaker 2: and there are big, big concerns. Our textbit poor Stenhouse

5
00:00:18,533 --> 00:00:20,973
Speaker 2: is here with the details. Good a Paul, Yeah, Good

6
00:00:21,013 --> 00:00:21,493
Speaker 2: morning Jack.

7
00:00:21,533 --> 00:00:24,333
Speaker 3: It's hard do sometimes know with this AI stuff when

8
00:00:24,373 --> 00:00:27,813
Speaker 3: we hear these developments, how much of it is sort

9
00:00:27,853 --> 00:00:30,893
Speaker 3: of the marketing height or the or you know, trying

10
00:00:30,933 --> 00:00:32,773
Speaker 3: to hype the investors. Do you give us some more

11
00:00:32,773 --> 00:00:35,173
Speaker 3: money and we need more money for infrastructure? And how

12
00:00:35,253 --> 00:00:37,933
Speaker 3: much of it is actually real? But I will say

13
00:00:37,973 --> 00:00:40,453
Speaker 3: that this model is coming from anthropic the ore, the

14
00:00:40,493 --> 00:00:43,693
Speaker 3: makers of Clawed and there I think it was their

15
00:00:43,773 --> 00:00:49,253
Speaker 3: CEO said that the fallout for economies, public safety, and

16
00:00:49,413 --> 00:00:53,973
Speaker 3: national security could be severe. So that kind of goes

17
00:00:54,013 --> 00:00:56,653
Speaker 3: beyond marketing hype, I think at that point. So this

18
00:00:56,693 --> 00:00:58,813
Speaker 3: is probably one we need to pay attention to.

19
00:00:59,653 --> 00:01:01,853
Speaker 2: I think that the way they've been discussing it, and

20
00:01:01,893 --> 00:01:03,613
Speaker 2: I love the stuff as you do, but I think

21
00:01:03,613 --> 00:01:06,053
Speaker 2: the way they've been discussing it suggests that they're seriously

22
00:01:06,053 --> 00:01:09,853
Speaker 2: concerned and about liability, Like they aren't talking it up

23
00:01:09,893 --> 00:01:11,693
Speaker 2: in a big way because they want to make money

24
00:01:11,733 --> 00:01:13,933
Speaker 2: out of it necessarily, though of course they do. It's

25
00:01:13,973 --> 00:01:16,493
Speaker 2: more that they're concerned this is so powerful that if

26
00:01:16,533 --> 00:01:19,893
Speaker 2: they release it for everyone overnight, the things that people

27
00:01:19,933 --> 00:01:22,813
Speaker 2: could potentially do with it would leave them liable for

28
00:01:22,933 --> 00:01:26,733
Speaker 2: lawsuits that would bankrupt the company, you know, I mean yeah, I.

29
00:01:26,653 --> 00:01:29,173
Speaker 3: Mean when you start talking to some of the things

30
00:01:29,173 --> 00:01:32,453
Speaker 3: here that it could be finding. Right, So this latest

31
00:01:32,453 --> 00:01:36,653
Speaker 3: model mythists is able to find. They're basically saying that

32
00:01:36,693 --> 00:01:39,213
Speaker 3: it is probably in the top one percent of software

33
00:01:39,213 --> 00:01:41,693
Speaker 3: engineers around. They're saying that it is in the top

34
00:01:41,733 --> 00:01:45,293
Speaker 3: one percent of cybersecurity professionals around. And so it is

35
00:01:45,373 --> 00:01:48,893
Speaker 3: able to find. It has found thousands, thousands of vulnerabilities

36
00:01:48,893 --> 00:01:52,973
Speaker 3: and software that we assumingly use day in and day out.

37
00:01:53,133 --> 00:01:55,533
Speaker 3: Some of the vulnerabilities that found have been there for

38
00:01:55,613 --> 00:01:57,573
Speaker 3: twenty seven years. No one knew they were there, even

39
00:01:57,613 --> 00:01:59,373
Speaker 3: the companies that have been testing it and know all

40
00:01:59,413 --> 00:02:02,493
Speaker 3: about the software. So this is like very powerful stuff.

41
00:02:02,533 --> 00:02:06,213
Speaker 3: And I think the thing that is probably most alarming

42
00:02:06,253 --> 00:02:09,613
Speaker 3: is it's not like it could just you know, create

43
00:02:09,653 --> 00:02:13,293
Speaker 3: a really cool app, you know that's for good. They're

44
00:02:13,333 --> 00:02:15,773
Speaker 3: really concerned about the bad And the thing that kind

45
00:02:15,813 --> 00:02:18,573
Speaker 3: of got me you know, set my alarm bells off.

46
00:02:18,693 --> 00:02:23,293
Speaker 3: Was the US Treasury Secretary. He actually in the wake

47
00:02:23,373 --> 00:02:25,693
Speaker 3: of this, a bunch of the US banks happened to

48
00:02:25,693 --> 00:02:28,413
Speaker 3: be in Washington for some lobbying type of things. But

49
00:02:28,893 --> 00:02:31,693
Speaker 3: he met with all of these critical banks. And you

50
00:02:31,733 --> 00:02:34,413
Speaker 3: can imagine, or I guess you don't want to imagine,

51
00:02:34,413 --> 00:02:37,453
Speaker 3: but you could be a time where imagine if they

52
00:02:37,453 --> 00:02:40,933
Speaker 3: could get into the kind of the critical big US

53
00:02:41,093 --> 00:02:44,773
Speaker 3: banks and they could start messing with transactions or messing with,

54
00:02:45,253 --> 00:02:48,093
Speaker 3: you know, the flow of money like that would just

55
00:02:48,373 --> 00:02:51,053
Speaker 3: upend them. So that would it would be and so

56
00:02:51,133 --> 00:02:54,333
Speaker 3: that wouldn't just be a lawsuit and some liability, that

57
00:02:54,373 --> 00:02:56,013
Speaker 3: would be something quite different.

58
00:02:56,253 --> 00:02:59,613
Speaker 2: And this is the concern, right, So they've basically tested

59
00:02:59,653 --> 00:03:04,053
Speaker 2: this model. They've asked this model to find cybersecurity vulnerabilities

60
00:03:04,573 --> 00:03:07,533
Speaker 2: and some of the most important software in the world.

61
00:03:07,733 --> 00:03:11,373
Speaker 2: And so this is stuff like internet browsers. It's the

62
00:03:12,253 --> 00:03:14,413
Speaker 2: little bits of software that none of us even know

63
00:03:14,813 --> 00:03:17,373
Speaker 2: is being used. So the open source stuff that gets

64
00:03:17,493 --> 00:03:20,493
Speaker 2: used in all manner of software that you use every

65
00:03:20,573 --> 00:03:22,973
Speaker 2: day without even kind of being aware that you're using

66
00:03:22,973 --> 00:03:25,933
Speaker 2: software at a time. And you know, there are examples

67
00:03:25,973 --> 00:03:29,573
Speaker 2: that I've read about where, for example, a piece of

68
00:03:29,613 --> 00:03:35,573
Speaker 2: software has been checked five million times by human cybersecurity experts,

69
00:03:35,853 --> 00:03:40,653
Speaker 2: coding experts, and artificial intelligence models and has never been

70
00:03:40,653 --> 00:03:42,893
Speaker 2: found to have a problem. And now they've put this

71
00:03:42,973 --> 00:03:46,333
Speaker 2: mythos model on it and it's immediately found massive vulnerabilities

72
00:03:46,373 --> 00:03:50,573
Speaker 2: that could be catastrophic. And it's this kind of example replicated, however,

73
00:03:50,653 --> 00:03:53,813
Speaker 2: many times over with However, many companies, however, many forms

74
00:03:53,853 --> 00:03:55,453
Speaker 2: of software that is really worrying.

75
00:03:56,133 --> 00:03:57,893
Speaker 3: Yes, And one of the ones that also got my

76
00:03:58,053 --> 00:03:59,693
Speaker 3: kind of hairs on the back of my neck up

77
00:03:59,773 --> 00:04:02,573
Speaker 3: was Cisco. So the CEO of Cisco, and Cisco makes

78
00:04:02,693 --> 00:04:06,813
Speaker 3: networking kind of hardware, right, but obviously behind the hardware

79
00:04:06,853 --> 00:04:09,253
Speaker 3: is also some software. Now, when you've got the CEO

80
00:04:09,293 --> 00:04:12,373
Speaker 3: of Cisco coming and saying that AI has reached a

81
00:04:12,453 --> 00:04:17,453
Speaker 3: level where critical infrastructure needs to be protected, as Cisco,

82
00:04:17,693 --> 00:04:22,533
Speaker 3: who provides the critical infrastructure for many digital businesses and

83
00:04:23,133 --> 00:04:27,453
Speaker 3: huge networks, you have to think we might be on that,

84
00:04:27,533 --> 00:04:29,853
Speaker 3: we might be at a time where we might need

85
00:04:29,853 --> 00:04:30,573
Speaker 3: to pump the brakes.

86
00:04:30,653 --> 00:04:33,733
Speaker 2: Yeah yeah, so interesting A. So one of the steps

87
00:04:33,733 --> 00:04:36,693
Speaker 2: this company has taken is they've got an alliance of

88
00:04:36,733 --> 00:04:42,253
Speaker 2: cybersecurity companies and like another three or four dozen organizations,

89
00:04:42,253 --> 00:04:46,173
Speaker 2: and they've given them like one hundred million dollars worth

90
00:04:46,213 --> 00:04:48,333
Speaker 2: of tokens to use this model so they can go

91
00:04:48,373 --> 00:04:50,133
Speaker 2: and test it before it's released in the public, which

92
00:04:50,133 --> 00:04:51,933
Speaker 2: seems like probably a very good idea, and they may

93
00:04:51,973 --> 00:04:53,733
Speaker 2: need more, who knows, but they'll be fascinating to see

94
00:04:53,773 --> 00:04:56,493
Speaker 2: what happens with this. Thank you, Paul. That is Paul Stenhouse,

95
00:04:56,653 --> 00:04:59,773
Speaker 2: our texpert this morning. Mythos is the name of that

96
00:04:59,853 --> 00:05:02,573
Speaker 2: new model from Anthropic. It has not been released to

97
00:05:02,573 --> 00:05:04,573
Speaker 2: the general public and it might be some time before

98
00:05:04,573 --> 00:05:04,813
Speaker 2: it is.

99
00:05:05,653 --> 00:05:08,773
Speaker 1: For more from Saturday Morning with Jack Tame, listen live

100
00:05:08,853 --> 00:05:11,693
Speaker 1: to News Talks at B from nine am Saturday, or

101
00:05:11,733 --> 00:05:13,653
Speaker 1: follow the podcast on iHeartRadio