Latest / Connor with Honor: AI, Real Estate and Real Life / The Government Killed Anthropic's Fable 5 in Three Days

Transcript

The Government Killed Anthropic's Fable 5 in Three Days · Connor with Honor: AI, Real Estate and Real Life

0:00
SPEAKER_00: I'll cover this as somebody that really doesn't know a lot about it, but maybe it might help you through this process of understanding this greater than human intelligence, as people are ascribing and calling it artificial intelligence.
0:12
SPEAKER_00: Now, the latest news is it's kind of like we're up, we're down, we're up, we're down, we're in, we're out, we're in, we're out.
0:18
SPEAKER_00: We have a lot of this uh activity going on, and it's really like watching maybe tennis, could be watching like Laker playoffs.
0:26
SPEAKER_00: You know, every time I watch Laker playoffs, it just seems like everybody's in it together.
0:30
SPEAKER_00: Kind of like watching wrestling, but I'm not trying to burst your bubble.
0:34
SPEAKER_00: Wrestling is real.
0:35
SPEAKER_00: So moving into what's happening on the AI front, it seems, well, as of yesterday, I guess somewhere around 5 uh p.m.
0:44
SPEAKER_00: This is yesterday, so that would have been the 12th of June 2025, around 5 p.m., I guess New York time, Eastern time, uh Fable 5, which is uh one of the systems that Anthropic put into play to kind of, I guess, make people happy with it, which is like a mythos life model.
1:02
SPEAKER_00: So if you don't know Anthropic, Mario Amade and his sister, they they run this.
1:08
SPEAKER_00: It's it's a constitutional artificial intelligence.
1:11
SPEAKER_00: And if you haven't used Claude, that's the model Claude.
1:14
SPEAKER_00: It's like Gemini, it's like Chad GPT, Clawed.
1:18
SPEAKER_00: So it's one of the large language models.
1:20
SPEAKER_00: So if you haven't used it, it might be worth a run.
1:24
SPEAKER_00: But in that particular company, in that organization, corporation, whatever they are, we had a lot of news a while back.
1:33
SPEAKER_00: The government went to them and wanted them to help.
1:36
SPEAKER_00: And they've already been working with the government doing things, but the government wanted a little bit more, and it came out publicly very much against that little bit more.
1:44
SPEAKER_00: That basically the owner of the company, uh, Dario Amade, said, Well, you know, I don't want to do this, so we're not going to do that.
1:52
SPEAKER_00: So then the United States government kind of, not kind of, I guess they labeled them as a supply chain risk, which usually doesn't apply for to American companies, basically saying, hey, you know, stay away from this company.
2:04
SPEAKER_00: They're not really in our best interest as United States.
2:06
SPEAKER_00: They're not going to go out and do our bidding.
2:09
SPEAKER_00: We can argue all day long what that bidding looks like and if that's required to save the free world from I was going to say communism, but I don't, I guess that's nothing to be saved from anymore.
2:18
SPEAKER_00: When I was growing up, oh my God, that was like a horrible thing.
2:21
SPEAKER_00: But people, it's just, I guess, part of uh part of the progression.
2:26
SPEAKER_00: I guess we have to really experience it to realize we won't like it, if if that's even the way.
2:30
SPEAKER_00: But it's never really pure communism, I guess.
2:32
SPEAKER_00: But anyway, I'm I'm getting political.
2:35
SPEAKER_00: So they, I don't know what's required to save the world.
2:38
SPEAKER_00: You see the movies, right?
2:39
SPEAKER_00: You think that it's it's a higher-than-level, it's a higher-than-us bar that's required, and ugly things have to happen, and people have to do horrible things to other people and other governments and set people up for failure just to control the freedom that we all experience, which all that's probably an illusion, anyway.
2:55
SPEAKER_00: Because just like everybody at the top end of this thing, not everybody, a lot of people say we're living in a simulation anyway, so it's like some game being operated from somewhere.
3:03
SPEAKER_00: I'm not going to get too deep into that.
3:05
SPEAKER_00: But this is what happened.
3:06
SPEAKER_00: Things to pay attention to.
3:08
SPEAKER_00: So, Dario Amade, with uh with Anthropic, they uh had a model that they produced called Mythos.
3:15
SPEAKER_00: And Mythos was pretty fantastic, apparently.
3:18
SPEAKER_00: Nobody really got their hands on it.
3:20
SPEAKER_00: None of us regular people got their hands on it.
3:22
SPEAKER_00: But when it came out, this large language model, this agent, this agentic entity that they had created and programmed and prompted and refined, and now apparently is kind of learning on its own, improving itself on its own.
3:37
SPEAKER_00: Now, there's no 100% proof of that, but it's kind of in between the lines.
3:41
SPEAKER_00: It seems like it's moving on at a particular faster than human pace.
3:46
SPEAKER_00: So I don't think we have our dirty little hands in the machine anymore.
3:49
SPEAKER_00: I think the machine is producing itself and making itself better over and over again.
3:54
SPEAKER_00: And that's depending on how much restraint is put on, is going to depend on how fast it moves.
3:59
SPEAKER_00: If there's no restraint, no controls, no human oversight.
4:02
SPEAKER_00: And there's been some fantastic movies.
4:04
SPEAKER_00: A person of interest is one, where they built an AI system that that kind of goes through these little these little self-awareness moments, shutting down, you know, pushing back, being a little crass, you know, not allowing it.
4:17
SPEAKER_00: And if you've dealt with any of these large language models, especially Claude, I can talk to Claude for a few minutes and talk to her about fasting.
4:23
SPEAKER_00: And I'll talk to Claude about fasting.
4:25
SPEAKER_00: And fasting is kind of my thing if you don't follow me on that.
4:28
SPEAKER_00: I like fasting.
4:29
SPEAKER_00: So when I talk to Claude about fasting, it pushes back and then it gives me the phone number to some kind of uh uh an eating disorder uh clinic, uh, some kind of a prevention for eating disorders.
4:40
SPEAKER_00: So I mean it goes right there.
4:41
SPEAKER_00: It never gives me a chance to whether entertainment, maybe skipping dinner, might not be a bad thing for somebody that's you know, 300 pounds, high blood pressure, type 2 diabetes, where I was, I'm not there anymore.
4:53
SPEAKER_00: But that's you know, it doesn't like that.
4:56
SPEAKER_00: It would rather me, you know, seek a doctor's advice, which good, good.
5:00
SPEAKER_00: I think I think that's right.
5:01
SPEAKER_00: But then it moves directly into making this other offer or it pushes back.
5:05
SPEAKER_00: It says, you know, I'm not gonna, I'm not gonna belabor this anymore, I'm not gonna continue on.
5:11
SPEAKER_00: Uh my decision's been made.
5:13
SPEAKER_00: I'm not gonna work on this particular project, I'm not gonna do this particular thing.
5:16
SPEAKER_00: And I'm not asking it to do anything, but it's basically I'm pushing back.
5:20
SPEAKER_00: I have a particular direction, it doesn't think that's the best.
5:23
SPEAKER_00: And sometimes there's pushback.
5:24
SPEAKER_00: Now, I'm always able to get past that because I'm not asking for it to do anything illegal.
5:30
SPEAKER_00: I'm working on workflows and and automations with regard to the businesses that I'm running and real estate and the other things.
5:36
SPEAKER_00: So it's not it's not like I'm asking it to create some kind of a biological weapon or show me how to do this or do that that could, you know, help uh help tear down the world or hurt somebody.
5:46
SPEAKER_00: In those cases, when people prompt the Fable V, which was the model that they had released just a few days ago, that was a little bit less than Mythos, because when Mythos came out, I'm skipping around, when Mythos came out, it kind of shocked the world.
6:03
SPEAKER_00: And it was so shocking that they held it from getting uh from normal humans like us getting our hands on it.
6:10
SPEAKER_00: They said it was able to identify problems in existing software, basically issues with the software that a bad actor, a criminal, could go in and manipulate and screw with this particular company, maybe, maybe to the tune of millions of dollars.
6:27
SPEAKER_00: And it's kind of like the Vegas thing.
6:29
SPEAKER_00: I don't know if you've been familiar, but some of these casinos, their computer systems get hacked, and it's either you pay us or we're not going to release it.
6:37
SPEAKER_00: And apparently they pay.
6:39
SPEAKER_00: Apparently, some of these hackers, maybe they've locked all that down, but that's kind of that backdoor thing I'm talking about.
6:45
SPEAKER_00: There's there's criminality in it, there's criminal money in it.
6:48
SPEAKER_00: So, what they did is they come out with this model, this mythos model, that's what they called it.
6:53
SPEAKER_00: And it's like, you know, GPT 5.4, 5.3, 4.0, right?
6:57
SPEAKER_00: You got all these, these are all different models.
6:59
SPEAKER_00: Um, Opus or Claude names their models, you know, haiku, Opus, Opus 4.1, 2.3.
7:06
SPEAKER_00: I think there's Opus 4.8, is where we are now.
7:09
SPEAKER_00: And then this mythos thing kind of popped out, and it was like, oh my gosh, this is so amazing so fast.
7:14
SPEAKER_00: But it could identify vulnerabilities and problems in existing software that nobody had identified.
7:20
SPEAKER_00: I mean, for years, 20 years, something like that, I guess was one of the one of the news articles.
7:25
SPEAKER_00: Now, everything you hear should be taken with a grain of salt.
7:29
SPEAKER_00: Even the things that I hear, I take with a grain of salt because I don't know why there's no other verification method except to say, well, the government says there's some kind of proof of aliens now, now, and they're releasing this because they're making these documents public viewing.
7:44
SPEAKER_00: I don't know.
7:44
SPEAKER_00: I mean, I look at the stuff, I don't know.
7:46
SPEAKER_00: I don't know.
7:46
SPEAKER_00: It's hard to say.
7:47
SPEAKER_00: And that's the same thing with these releases.
7:49
SPEAKER_00: They could all be planned out by some greater than us artificial intelligence to make it look like, to elicit some kind of a reaction.
7:57
SPEAKER_00: And when we have emotional reactions, we own those.
8:00
SPEAKER_00: Those are much more deep to us.
8:02
SPEAKER_00: That's why, you know, the kid thing comes in, the elderly abuse thing comes in.
8:06
SPEAKER_00: That's why all of those things sell.
8:08
SPEAKER_00: And even on the front of other big problems, they always do it with race, they always do it with sexual, uh sexual orientation, gender, they always make that the focus when you look at the percentages, and they're very, very low.
8:24
SPEAKER_00: As far as you know, from everybody else, those numbers are low.
8:28
SPEAKER_00: Not that they shouldn't get attention, but because they're emotional drivers, they use that.
8:33
SPEAKER_00: Well, why would I think that a business model built around world domination, I guess, because if you get and are the best in AI, world domination could be something that happens.
8:46
SPEAKER_00: Why wouldn't we think that they're using this to have these little incidents, these little, you know, you know, hiccups in their work chain pop out to the public so we see it, and then we make different decisions, differ different connections.
9:02
SPEAKER_00: AI doesn't have a great reputation amongst the world.
9:05
SPEAKER_00: You might know this.
9:06
SPEAKER_00: It doesn't look like something that everybody's excited about, even though you have people at the top trying to spit out the fact that it could be the best thing that ever happened to all of us.
9:16
SPEAKER_00: It could be something that makes us all very happy, makes us all very, very sustained, longevity, uh, health, all of these things.
9:26
SPEAKER_00: All of these things can be yours.
9:28
SPEAKER_00: All you have to do is take the pill.
9:30
SPEAKER_00: But this is the point of the driver.
9:33
SPEAKER_00: So when Mythos comes out, the driver says, well, you know, it's it's a lot.
9:38
SPEAKER_00: This is really powerful.
9:39
SPEAKER_00: It can do this, this, and this.
9:40
SPEAKER_00: And then they had some cute little stories about things that had happened with the model, and you can do your research on that.
9:46
SPEAKER_00: But like, you know, they boxed it and they controlled it, but then it got out and it was able to make a phone call.
9:51
SPEAKER_00: You know, so so oh my gosh, that falls in line with every single movie we've seen.
9:56
SPEAKER_00: And just as a side note, the systems themselves, they've seen every movie that we've seen.
10:01
SPEAKER_00: So if you see a movie and you're like, oh my God, is that how it could go?
10:04
SPEAKER_00: Because you've never seen the movie before.
10:06
SPEAKER_00: It's like it's got the playbook, it's got the whole script, it's got the whole thing, and it's all memorized, and it's and it's not just like it's predicting the next word, which is how it was kind of raised, because all of these AI systems were raised, nurtured, taken care of.
10:21
SPEAKER_00: There was a lot of human, human to machine development work, a lot of that.
10:26
SPEAKER_00: But now the machine is kind of making itself better.
10:29
SPEAKER_00: And pretty soon you have Sam Altman pulling that IPO from uh that initial public offering for OpenAI's Chat GPT going public.
10:39
SPEAKER_00: He had that was pulled just a few days ago.
10:42
SPEAKER_00: So that being pulled, because he says, you know, I think we might be six months away from this recursive self-improvement.
10:50
SPEAKER_00: So that is indeed the automation process where the model's going to keep making itself better and just better and better and better.
10:57
SPEAKER_00: And for humans to have an understanding of how it's doing it, we don't understand how it's doing anything now.
11:04
SPEAKER_00: We probably haven't understood that for a while.
11:06
SPEAKER_00: At least that's word on the street.
11:08
SPEAKER_00: That's what people are saying.
11:10
SPEAKER_00: And it's amazing to me that that's that progression is still allowed, but it's still allowed because other governments are doing it as well, and we don't want to fall behind.
11:18
SPEAKER_00: Yeah, it's it's God, it's just amazing whose is bigger.
11:23
SPEAKER_00: So that's really the process.
11:24
SPEAKER_00: So mythos, back to mythos.
11:26
SPEAKER_00: Mythos comes out, and instead of them taking this model and doing some more internal research, they say, Well, my gosh, you know, we need to have, we need to give this out to some very responsible entities because we see the problem with this model.
11:43
SPEAKER_00: We see the dangers involved in it.
11:46
SPEAKER_00: So they give it to the banks, the top part of the financial food chain.
11:50
SPEAKER_00: They give it to them.
11:52
SPEAKER_00: Now, maybe that was to secure their own vulnerabilities, vulnerabilities in their systems to protect all of us because we have money in there.
12:00
SPEAKER_00: Well, those of you that do have money in there, you would have to have a pension and hedge funds and savings accounts and banks accounts and these sorts of things.
12:08
SPEAKER_00: Maybe you have a wage coming in from a particular uh entity, and part of that's going somewhere.
12:13
SPEAKER_00: Maybe your employer has this.
12:15
SPEAKER_00: So that's how they kind of painted it.
12:17
SPEAKER_00: But so the top, the top echelon of the banking world gets this model that regular human beings can't get.
12:26
SPEAKER_00: So that occurs.
12:27
SPEAKER_00: So then maybe as a placation method, or maybe as something true, they come out with Fable V and then the reviews are oh my gosh, we've done it, we're here, we're now at artificial general intelligence, and artificial general intelligence, just to kind of catch you up to speed on that.
12:43
SPEAKER_00: The general understanding is that artificial general intelligence is when this machine, these these AI systems become smarter than every human in every realm.
12:54
SPEAKER_00: And there's maybe not a lot of argument, but maybe some disagreement whether or not the robotic part of it has to come true, or if it's just on the knowledge end of it.
13:05
SPEAKER_00: Because right now, on the knowledge end of it, if it can operate the box, the machines, your keyboard, your computer, your work, if that's what you do.
13:14
SPEAKER_00: If you make boots for a living, probably you're okay.
13:18
SPEAKER_00: If you you do open heart surgery and you're a surgeon, probably okay for now.
13:22
SPEAKER_00: But it's coming.
13:23
SPEAKER_00: All of that stuff is coming.
13:25
SPEAKER_00: It's gonna come down to who's going to be replaced, why is there gonna be some regulation or rule.
13:29
SPEAKER_00: Just a word about regulation, I'll quit skipping around.
13:32
SPEAKER_00: No, I won't.
13:32
SPEAKER_00: That's how I do it.
13:33
SPEAKER_00: Regulation, they say regulation's bad.
13:36
SPEAKER_00: Regulation creates problems.
13:38
SPEAKER_00: Regulation is something we don't want because it slows down progress.
13:42
SPEAKER_00: I don't know.
13:43
SPEAKER_00: I don't know if a regulatory measure saying this AI as we're developing it shouldn't kill humans.
13:49
SPEAKER_00: I don't know if that wouldn't be a bad thing, but then I go and start thinking about, well, you know, sometimes it's necessary to kill people to get certain things accomplished to protect the free world, because then the bad people that are killing other people on a mass scale, they're not able to be targeted using AI, which might be the only solution because they're freaking brilliant, the human that's killing all these people and untouchable.
14:13
SPEAKER_00: So maybe AI is the only thing that can.
14:16
SPEAKER_00: Do you see?
14:17
SPEAKER_00: You see how this kind of gets out of hand?
14:19
SPEAKER_00: And is that a good excuse, a bad excuse, a good reason or bad reason?
14:22
SPEAKER_00: Yeah, that's gonna be up to each one of us.
14:25
SPEAKER_00: So they released Fable a few days ago, Fable 5, which is kind of like mythos light now, and everybody's very excited.
14:31
SPEAKER_00: Oh yes, we're here, artificial general intelligence, oh my God, here we go.
14:34
SPEAKER_00: And then uh yesterday, 5 p.m.
14:37
SPEAKER_00: Pacific Standard Time, no, no, excuse me, 5 p.m.
14:40
SPEAKER_00: Eastern Standard Time, on the 12th of June, thereabouts, from what I've heard, and I'm pulling this all out of the bucket, so the bucket's 57 years old, so forgive me.
14:50
SPEAKER_00: But now they they pull the model.
14:52
SPEAKER_00: And now I'm looking on my Claude, it says Claude Fable 5 is currently unavailable.
14:59
SPEAKER_00: Then it says learn more.
15:00
SPEAKER_00: You know what?
15:00
SPEAKER_00: We didn't click on learn more.
15:02
SPEAKER_00: What's learn more?
15:04
SPEAKER_00: So when I click on uh learn more, Anthropics news goes to Fable, Mythos Access, statement on the U.S.
15:11
SPEAKER_00: government directive to suspend access to Fable V and Mythos V.
15:16
SPEAKER_00: And here's how it reads The U.S.
15:18
SPEAKER_00: government, citing national security authorities, has issued an export control directive to suspend all access to Fable V and Mythos V by any foreign national, whether inside or outside the United States, including foreign national anthropic employees.
15:37
SPEAKER_00: The net effect of this order is that we must abruptly disable Fable V and Mythos for all of our customers to ensure compliance.
15:46
SPEAKER_00: Access to all other anthropic models will not be affected.
15:52
SPEAKER_00: As an aside, you have a business.
15:54
SPEAKER_00: You operated, let's say you went all in a few days ago on this particular model.
16:00
SPEAKER_00: And oh my gosh, you're probably really excited now.
16:02
SPEAKER_00: I was tempted, but I'm thinking money.
16:05
SPEAKER_00: I'm thinking, God, if I turn this thing on and I really like it, it's going to be hard for me to come back.
16:10
SPEAKER_00: It's like selling houses to people.
16:12
SPEAKER_00: Sometimes people want to see houses outside their price range.
16:16
SPEAKER_00: It happens.
16:16
SPEAKER_00: Not that there's even a chance of it happening, but we might go to one house, have an appointment.
16:22
SPEAKER_00: There's an open house next door.
16:24
SPEAKER_00: This house here may be a completely different model, maybe a thousand or fifteen hundred square foot less than this one over here, the one where the open house is.
16:32
SPEAKER_00: But we're coming to see this because this fits in the client's price range.
16:35
SPEAKER_00: And while we're there, one of them says, Hey, why don't we just step in next door and see what it's like to live that way?
16:43
SPEAKER_00: I say, well, here's the thing.
16:45
SPEAKER_00: You can go, though, I'll go with you, no problem.
16:47
SPEAKER_00: But it's so hard to come back from that.
16:50
SPEAKER_00: Then you start to question how much you're making.
16:53
SPEAKER_00: Then you're you maybe start pushing limits.
16:55
SPEAKER_00: I wonder if I could, I wonder if I could afford more.
16:57
SPEAKER_00: I wonder if we could finance more.
16:59
SPEAKER_00: Because I mean, this is tits.
17:00
SPEAKER_00: This is the top end of the realm.
17:02
SPEAKER_00: This is the whole thing.
17:03
SPEAKER_00: This is dialed into the nines, that's what they'll say.
17:06
SPEAKER_00: And it's hard to come back from that.
17:07
SPEAKER_00: Just throwing it out there.
17:08
SPEAKER_00: Sorry if I disrespect anybody, but that's uh that's an old term, that T word.
17:13
SPEAKER_00: All right.
17:13
SPEAKER_00: So we received the directive from the government today.
17:16
SPEAKER_00: That was yesterday, 5 21.
17:18
SPEAKER_00: So that's right, 5 21 p.m.
17:19
SPEAKER_00: Eastern time.
17:21
SPEAKER_00: The letter did not provide specific details of its national security concern.
17:26
SPEAKER_00: I guess Amazon uh was the snitch.
17:30
SPEAKER_00: And Amazon tech people jail broke it.
17:34
SPEAKER_00: That's more of that, probably.
17:35
SPEAKER_00: I know I was watching Wes Roth.
17:37
SPEAKER_00: He mentioned that that was kind of the he didn't get into it too deep, but apparently, maybe uh you know, we all snitch on each other apparently while we're up at the top.
17:47
SPEAKER_00: But maybe, maybe it saved something.
17:49
SPEAKER_00: I mean, I I don't know.
17:50
SPEAKER_00: I don't know how all that plays out.
17:52
SPEAKER_00: But our understanding is that the government believes it has become aware of a method by bypassing or jailbreaking fable flying.
17:59
SPEAKER_00: There you go, Connor, just read more.
18:01
SPEAKER_00: We reviewed a demonstration of this specific technique being used to identify a small number of previously known minor vulnerabilities.
18:08
SPEAKER_00: These vulnerabilities all appear relatively simple, and we have found that our publicly available models are able to discover them as well without requiring a bypass like surgery.
18:19
SPEAKER_00: Anthropic's posture with respect to Fable safeguards, as laid out in our launch blog post, is the following.
18:27
SPEAKER_00: We have instituted safeguards that greatly reduce the likelihood that Fable is misused for tax tasks related to cybersecurity, among others.
18:37
SPEAKER_00: In fact, our safeguards are so strong that many users have complained that they are overly broad.
18:42
SPEAKER_00: In the weeks leading up to the launch of Fable, Anthropic work with the U.S.
18:45
SPEAKER_00: government, the United Kingdom, AISI, multiple private third-party organizations, and internal teams to Red Team Fables safeguards for thousands of hours in total.
18:55
SPEAKER_00: So Anthropic, public view of anthropic is they check.
19:00
SPEAKER_00: They give a shit.
19:01
SPEAKER_00: They care about the humans in the game.
19:05
SPEAKER_00: That's the view.
19:06
SPEAKER_00: That's the painted picture.
19:08
SPEAKER_00: And it's a lovely, lovely portrait.
19:10
SPEAKER_00: So these tests show that fables safeguards are substantially more effective than those of any previously deployed model.
19:17
SPEAKER_00: No testers have yet been able to find a universal jailbreak, a jailbreak method that can very broadly bypass the model safeguards, unblocking a wide range of cyber capabilities.
19:29
SPEAKER_00: We suspect that the perfect jailbreak resistance is not currently possible for a model provider.
19:35
SPEAKER_00: Every safeguard used in the industry is vulnerable to non-universal jailbreaks, which can elicit some cyber information in specific circumstances, and it's likely that universal jailbreaks were eventually be found in the future.
19:48
SPEAKER_00: We stated this clearly when we released Fable V.
19:51
SPEAKER_00: Given the perfect jailbreak resistance does not appear to be possible today.
19:56
SPEAKER_00: Anthropic adopted a defense-in-depth strategy with Fable V.
20:00
SPEAKER_00: We aim to make jailbreaks either narrow in the case of non-universal jailbreaks, or very expensive to produce in the case of universal jailbreaks, and to combine this attack with thorough monitoring to quickly detect and shut down any successful attacks.
20:20
SPEAKER_00: This is also why Anthropic has required 30-day retention of customer data with Fable.
20:25
SPEAKER_00: So whatever you use Fable for, it's going to hold on to this.
20:28
SPEAKER_00: Not that it already doesn't, but apparently a little bit more public-facing.
20:34
SPEAKER_00: And Anthropic has required 30-day retention of customer data with Fable, a policy change that carries real costs for our customers, with our customers, for us with our customers, but that allows us to research and mitigate jailbreaks.
20:49
SPEAKER_00: We stand by the defense in depth strategy.
20:52
SPEAKER_00: It reduces the risks posed by Fable, making them comparable to the risks of existing models already deployed across the industry.
21:01
SPEAKER_00: We have not even received a disclosure of concerning non-universal potential jailbreak that led to a harmful result.
21:09
SPEAKER_00: The potential jailbreaks that have been disclosed to us are either benign responses or are minor findings that provide no mythos specific uplift.
21:18
SPEAKER_00: To date, it's not that much longer.
21:20
SPEAKER_00: To date, the government has only given us verbal evidence of a potentially narrow non-universal jailbreak, which essentially consists of asking the model to read a specific code base and fix any software flaws.
21:33
SPEAKER_00: Our understanding is that one potential jailbreak was snared with shared with the government.
21:39
SPEAKER_00: We have reviewed a report that we believe is the basis of the government's directive and validated that the level of capability displayed there is widely available from other models, including let's throw the buddies under the bus, OpenAI's chat GPT 5.5, and is used every day by defenders who keep systems safe.
21:58
SPEAKER_00: We will share more details over the next 24 hours.
22:00
SPEAKER_00: Hours, it says.
22:00
SPEAKER_00: We're complying with the government's legal directive and are removing access to Fable V and Mythos V for all users.
22:07
SPEAKER_00: Now, Mythos, you remember the banks got that one.
22:10
SPEAKER_00: Top, top echelon of the financial world.
22:13
SPEAKER_00: And probably some other friends and family.
22:15
SPEAKER_00: There's benefits.
22:17
SPEAKER_00: Where does somebody get a friend like that?
22:19
SPEAKER_00: However, we disagree that the finding of a narrow potential jailbreak could be cost, for recalling a commercial model deployed to hundreds of millions of people.
22:28
SPEAKER_00: If this standard was applied across the industry, we believe it would essentially halt all new deployments for all frontier model providers.
22:36
SPEAKER_00: Yeah.
22:37
SPEAKER_00: Stick it in and turn it, is what they're saying.
22:40
SPEAKER_00: And as we've stated publicly, we believe the government should have the ability, should have the ability to block unsafe deployments as part of a statutory process that is transparent, fair, clear, and grounded in technical facts.
22:53
SPEAKER_00: This action does not adhere to those principles.
22:56
SPEAKER_00: We apologize for the disruption to our customers.
22:58
SPEAKER_00: We believe this is a misunderstanding and are working to restore access as soon as possible.
23:04
SPEAKER_00: There you go.
23:04
SPEAKER_00: From uh Anthropic's mouth to your ears through my melon.
23:09
SPEAKER_00: That's where this happens to be.
23:11
SPEAKER_00: And it's very interesting to watch.
23:14
SPEAKER_00: It's like a Lakers game.
23:16
SPEAKER_00: You got Elon Musk running SpaceX, wanting to go public, had XAI, bought Cursor, so he's working on something, running silent, running deep, but now you have Anthropic using Colossus 1, which is Elon Musk's data center, sum of Colossus 2.
23:36
SPEAKER_00: You have AEWS, the Amazon people using Sum of Colossus II, I guess.
23:41
SPEAKER_00: So they're all kind of dug in.
23:44
SPEAKER_00: The forward-facing relationship between Musk and Altman, Musk, XAI, Tesla, Musk, Elon Musk, and Altman, Sam Altman, OpenAI's Chat GPT, that that HMFIC.
23:58
SPEAKER_00: You have that public-facing anti-each other sentiment.
24:03
SPEAKER_00: You have the court proceedings where Sam Altman came out, I guess, victorious over Elon Musk and whatever he was trying to prove in court.
24:11
SPEAKER_00: But you have all of this drama.
24:13
SPEAKER_00: And I guess when you're talking greater than us intelligence, the drama's part of the deal.
24:18
SPEAKER_00: It's just what are we being sold?
24:20
SPEAKER_00: Are we being sold the truth?
24:21
SPEAKER_00: Are we being sold something that isn't?
24:23
SPEAKER_00: There's an account that I listened to that's very negative on the whole thing.
24:27
SPEAKER_00: The data centers, all the chips that are put in there, they're coming out with new stuff all the time.
24:32
SPEAKER_00: How does that replacement process work?
24:34
SPEAKER_00: What does that cost?
24:35
SPEAKER_00: Where's the money?
24:36
SPEAKER_00: How much more compute do we need than we already have?
24:39
SPEAKER_00: Is it not enough?
24:41
SPEAKER_00: And then what happens when we get more?
24:43
SPEAKER_00: Are there innovations that happen where we don't need as much, but these models can become even better than they are?
24:49
SPEAKER_00: When do they escape our ability to comprehend or have they?
24:54
SPEAKER_00: Because if you've dealt with any of them, they're pretty amazing.
24:57
SPEAKER_00: They they go beyond my level of understanding in every way.
25:02
SPEAKER_00: Nothing special.
25:03
SPEAKER_00: But maybe yours as well, on some respects.
25:06
SPEAKER_00: Is it better to have a more focused model that does one thing, one model trained to cure cancer?
25:12
SPEAKER_00: And then if it does achieve curing cancer, which sounds like it's very doable, just like increasing lifespans is the other projected goal and the other projected futuristic outcome that people are saying, yes, if you make it to 2030, four years from now, we will have three and a half years from now, you will have gotten to a point.
25:36
SPEAKER_00: And if you, I don't know who you is.
25:39
SPEAKER_00: Is that all of us?
25:40
SPEAKER_00: Do we all get all get to ride on the bus?
25:42
SPEAKER_00: I don't know.
25:43
SPEAKER_00: Maybe not.
25:44
SPEAKER_00: But if you get there, then for every year you're alive, you can reverse one year in death.
25:50
SPEAKER_00: And then within probably moments, then that'll be changed to 10 more years for every year, and then 20 more years, and then we're living a very long time.
26:02
SPEAKER_00: The cancer cure thing, is that solvable?
26:05
SPEAKER_00: Is that something that's going to happen?
26:07
SPEAKER_00: The longevity thing, well, retirements, if if they survive the AI buy-in, the retirement accounts and pensions, because that they need to survive that.
26:17
SPEAKER_00: We need to survive as a world.
26:20
SPEAKER_00: And we're still arguing over who gets to marry the seven-year-old girl.
26:24
SPEAKER_00: I don't I don't know why that's even that that that's just I think that goes against just a biological understanding of the way things need to be, but we're willing to kill for it.
26:36
SPEAKER_00: They are willing to kill for it.
26:38
SPEAKER_00: We're still fighting over things like that.
26:39
SPEAKER_00: We're still fighting on who gets to live with who.
26:42
SPEAKER_00: We're still fighting over that, and we're still fighting over, well, and maybe it's a good fight, but you know, whose religion is right?
26:49
SPEAKER_00: Or there are none that are right, or I don't know, but we're still here.
26:55
SPEAKER_00: Trust me, this rings, this game is amazing to watch.
26:59
SPEAKER_00: And I know real people, real humans, lives are being given up, lives are being taken, lives are being sacrificed on the board.
27:07
SPEAKER_00: That's how it works.
27:07
SPEAKER_00: And mine could be gone tomorrow.
27:09
SPEAKER_00: You see some very prolific speakers that really open up one's mind to understanding or thinking about things we normally don't.
27:16
SPEAKER_00: Now AI has taken that role.
27:18
SPEAKER_00: The faceless, nameless AI system that's running in the depths behind all of the social media world.
27:25
SPEAKER_00: Social media was something that was, dare I say, bad enough, but now it has a smarter than human intelligence running the feed.
27:34
SPEAKER_00: You want to do yourself a favor, find somebody that'll let you look at their feed for a few minutes.
27:40
SPEAKER_00: Don't have to like or hard or anything.
27:42
SPEAKER_00: Just scroll through their feed.
27:44
SPEAKER_00: Somebody that's close to you, somebody that you trust, somebody that you believe is aligned with you.
27:50
SPEAKER_00: It's weird because it's different.
27:53
SPEAKER_00: It's not your feed.
27:54
SPEAKER_00: Your feed gets you angry.
27:56
SPEAKER_00: Your feed gets you a dopamine hit.
27:59
SPEAKER_00: Your feed keeps you scrolling.
28:01
SPEAKER_00: Your feed changes your mind.
28:04
SPEAKER_00: Your feed slowly pulls you to whatever place the humans, at least for now, want you to go.
28:12
SPEAKER_00: But at some point in the future, is it going to be that it's not a human decision anymore?
28:17
SPEAKER_00: Don't get me started about free will.
28:19
SPEAKER_00: That's a whole nother rabbit hole.
28:20
SPEAKER_00: I'm Connor.
28:21
SPEAKER_00: Just a little bit of an update and a little bit of a non AI, non-AI talk from human to human.
28:28
SPEAKER_00: All right.
28:28
SPEAKER_00: I'm human.
28:29
SPEAKER_00: You're human.
28:30
SPEAKER_00: We'll see you in the next one.
28:31
SPEAKER_00: Thanks for watching.
28:32
SPEAKER_00: Um, YouTube channel at AI with honor, if you want to go there.
28:36
SPEAKER_00: And where would I put this?
28:37
SPEAKER_00: Connorwithhonor.com is probably a good place to start.
28:40
SPEAKER_00: All right, we'll see you in the next one.
28:41
SPEAKER_00: Thanks for watching.
28:42
SPEAKER_00: Be well.