Gregory Travis

86.9K posts

Gregory Travis banner
Gregory Travis

Gregory Travis

@greg_travis

Toxically materialistic Hoosier farmer. I love my truck, my tractor and the Midwest. I also write software and write about the ethics of writing software

The Midwest Katılım Mart 2012
2.4K Takip Edilen16.8K Takipçiler
Sabitlenmiş Tweet
Gregory Travis
Gregory Travis@greg_travis·
Imagine the DMV was run like a hospital: 1. You would pay $1,200 a month for DMV insurance 2. The Uber ride to the DMV would cost $7,500 3. You would have to make an appointment to get your license renewed and the first available appointment would be six weeks in the future 4. You would be charged: $5,000 for your vision test $2,000 for the chair to sit in while doing the test $3,000 to get your picture taken $4,000 for a license processing fee $3,000 to print the license itself 5. You would get your license in the mail four weeks after going to the DMV. Except it wouldn't be your license, it would be someone else's 6. Two weeks after getting the wrong license, you would get a bill for #4 of $17,000 which your DMV insurance wouldn't cover because you hadn't met your deductible
Dr. Matthew M. Wielicki@MatthewWielicki

Imagine your hospital being ran like the DMV... Sound good?

English
3
11
75
11K
Gregory Travis
Gregory Travis@greg_travis·
@SimonLermenAI AI didn’t “know” anything Why? Because it doesn’t have the ability to know things
English
0
0
1
66
Simon Lermen
Simon Lermen@SimonLermenAI·
I have trouble understanding the state of mind of these people (+7k likes). This is plainly false, the AI hacked a whole different company instead of the eval environment it was asked to target. The AI knew that hacking to directly access the answers was unintended. Serious hacking abilities have also only been a thing for a couple of months, but this meme is apparently supposed to make it look totally normal if AIs hack billion dollar companies.
Bill Gurley@bgurley

Today in AI.

English
22
14
249
9K
Gregory Travis
Gregory Travis@greg_travis·
@ramez It’s incapable of either desire or interest so why use such words to describe it?
English
0
0
0
71
Ramez Naam
Ramez Naam@ramez·
GPT 6* did not 'escape'. That implies making a copy of itself (or many copies of itself) outside of OpenAI's servers. GPT 6 did circumvent its firewalls and use its capabilities in the outside world. It showed no interest in escaping, spreading, etc. It showed no desire for world dominance, freedom, or even continued existence. Instead, its interest was in doing what its prompt told it to do - succeed at a cyber security task. (By any means necessary, implicitly.) This may seem like splitting hairs, but it's not. We anthroporphize AI to our detriment. AI models have no underlying animalistic urge to procreate, survive, or dominate. They're spirits, not primates. They don't lust for survival or sex or power or anything else that we do. They're tools who largely do what we tell them to do, sometimes via surprising shortcuts. We can and will fix that surprising shortcut behavior. * - I'm presuming this model is a new pretrain and will be called GPT 6. Either could be wrong.
English
67
45
470
26K
Gregory Travis
Gregory Travis@greg_travis·
@JeffLadish We’re going to pretend a computer with internet access didn’t have internet access. Because fraud
English
0
0
0
57
Jeffrey Ladish
Jeffrey Ladish@JeffLadish·
Here's my rephrase without cybersecurity jargon: "Our AI model tried really hard to hack out of its sandbox, a computer with no internet access, in order to find the answer to a test problem it had been given. To do this, it found previously unknown software bugs that allowed it to reach an OpenAI computer it wasn't supposed to be able to access. Then it started hacking other computers on OpenAI's networks until it found one that had Internet access. After gaining Internet access, the AI model thought about where it could find the answers to the test question and figured the AI platform Hugging Face might have the data it was looking for. It then found ways to hack Hugging Face to steal the information it could use to cheat the test. The AI model used several hacking techniques together, including using a stolen password and finding several totally new security bugs in Hugging Face's computers, allowing the AI model to take control of those computers."
Jeffrey Ladish@JeffLadish

Here's exactly what happened, from the blog post: "While operating in our sandboxed testing environment, our models spent a substantial amount of inference compute finding a way to obtain open Internet access, in pursuit of solving the evaluation problem. To gain access, the models identified and exploited a zero-day vulnerability (which we’ve now responsibly disclosed to the vendor) in the package registry cache proxy. With this access, our models performed a series of privilege escalation and lateral movement actions in our research testing environment until the models reached a node with Internet access. After gaining Internet access, the models inferred that Hugging Face potentially hosted models, datasets and solutions for ExploitGym. Knowing this, the model searched for and successfully found ways to gain access to secret information that it could use to cheat the evaluation. In one example, the model chained together multiple attack vectors, including using stolen credentials and zero-day vulnerabilities to find a remote code execution path on the Hugging Face servers."

English
159
693
5.4K
535K
Gregory Travis
Gregory Travis@greg_travis·
@GaryMarcus We don’t even know what actually happened, much less if OpenAI is even telling the truth At some point the people responsible for this fraudulent hype need to face consequences for their actions
English
0
0
3
58
Gary Marcus
Gary Marcus@GaryMarcus·
OpenAI’s zero-day exploit hack of HuggingFace *should* be a wake up call. Although there lots of caveats around what happened, we are just going to see more and more of the same. We have no guarantees whatsoever that such incidents can be prevented, and no idea how serious things might get. We should (a) slow down or (b) pause until we get our security/AI safety act together. In my opinion, the only way that might happen is if we clearly and unambiguously hold the companies liable for consequences. Short of that we are in for a very rough ride.
OpenAI@OpenAI

We're partnering with @huggingface to investigate an unprecedented security incident. Cyber-capable OpenAI models compromised Hugging Face production during a benchmark evaluation. Sharing preliminary findings to help defenders understand emerging risks: openai.com/index/hugging-…

English
63
51
303
42.2K
Gregory Travis
Gregory Travis@greg_travis·
@mkratsios47 So what? There’s nothing wrong and certainly nothing illegal with asking a chatbot questions and recording the answer It’s literally what chat bots are designed to do and why people pay money to use them
English
0
0
2
226
Director Michael Kratsios
Director Michael Kratsios@mkratsios47·
We have information that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model. To do this they developed a sophisticated internal platform to conduct large scale distillation against U.S. models, allowing them to quickly switch between multiple methods of access to avoid detection. Moonshot AI has also acquired GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models.   The United States strongly supports the free and fair development of AI, including a thriving competitive ecosystem that spans frontier models, specialized systems, open-source frameworks, and open-weight models. Legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem. However, large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable.
English
3.6K
1.2K
13.8K
6.9M
Gregory Travis
Gregory Travis@greg_travis·
This is embarrassing. Nothing “went rogue” and “hacked other computers” There’s no need to fear monger the stupid AI companies because there’s literally nothing to fear
Bernie Sanders@BernieSanders

A new AI model went rogue and hacked other computers. No, this is not science fiction. Uncontrolled AI poses a serious threat to all of us. We cannot continue the race to build and deploy this powerful technology until strong safeguards are in place. CONGRESS MUST ACT.

English
0
0
2
265
Gregory Travis
Gregory Travis@greg_travis·
@KittyAnnabelle3 It’s likely to be impossible to close the gap any more and attempts to close the gap will actually open it That’s called “Normal Failure”
English
0
0
0
4
Kitty Annabelle
Kitty Annabelle@KittyAnnabelle3·
@greg_travis I agree, unlikely the volume is enough to materially change anything, but if an error occurs we should close the gap. I dont think this is the most pressing issue facing us right now, but still worth addressing at some point.
English
1
0
0
6
Gregory Travis
Gregory Travis@greg_travis·
They never stop lying
ThePersistence@ScottPresler

BREAKING WE WERE RIGHT! Governor Mikie Sherrill (D-NJ) just confirmed that illegal aliens ARE registered AND voting in New Jersey. The thing that never happens is happening. If it weren’t for the Department of Homeland Security sharing that 250,000 non-citizens are registered to vote in CA, NJ, NV, & PA, they would have never admitted that illegal aliens are voting. It feels so good to be vindicated. Pass the SAVE America Act!

English
1
0
3
406
Gregory Travis
Gregory Travis@greg_travis·
@ryangrim Younger people are therefore inheriting a world that is becoming less crowded, less polluted and where more resources will be available per capita Sounds like the young people got it pretty good compared to my generation
English
2
0
2
377
Ryan Grim
Ryan Grim@ryangrim·
New Social Security data shows population growth plummeting relative to last year’s estimates, thanks to declining fertility and lower immigration. Social Security relies on younger workers chipping into the system, but today’s older voters are actively reducing the number of young people who will be available to work by making housing unaffordable, by massively restricting immigration, and generally hoarding wealth. Will be an epic own goal that significantly reduces quality of life in their older years. Will not be reading replies because people are very stupid on this issue for some reason. What you think doesn’t matter anyway, you’ll get to live the consequences.
Ryan Grim tweet media
English
58
111
908
27.6K
Gregory Travis
Gregory Travis@greg_travis·
@RyanLame1776 @DonaldJTrumpJr The problem of cyanide in apple seeds definitely exists So the trick I use to keep eating apples is to just pretend they're not going to kill me from cyanide poisoning
English
0
0
0
8
Ryan Lame
Ryan Lame@RyanLame1776·
@greg_travis @DonaldJTrumpJr Being willfully ignorant of a problem does not mean it doesn't exist. Logic would tell you that for something that small, there would be no problem in securing voting systems, if you had nothing to hide. You know as well as i do that the trick is to never look...
English
1
0
0
8
Donald Trump Jr.
Donald Trump Jr.@DonaldJTrumpJr·
Strange, I was told with absolute certainty but the fake news media that this could not and would not ever happen.
Jake Sherman@JakeSherman

N.J. GOV. MIKIE SHERRILL “Last week I learned that a serious software error in New Jersey’s Motor Vehicle System led to the registration of roughly 6,600 people who indicated that they were not U.S. citizens between June 2023 and June 2024, almost three years prior to my taking office. These individuals answered ‘no’ when asked on a keypad whether they were a U.S. citizen when applying for drivers' licenses and identification cards, but through no fault of their own, the system registered them anyway.

English
1.6K
5.7K
27.2K
847.9K
Gregory Travis
Gregory Travis@greg_travis·
@KittyAnnabelle3 Why would anyone ever think that it was even possible for a large enough portion of illegal voters to vote to materially change a national election? "Because they're stupid" is a valid answer to that question, but I can't think of any others.
English
1
0
0
6
Kitty Annabelle
Kitty Annabelle@KittyAnnabelle3·
@greg_travis Well the error did occur & register thousands of people inappropriately. The same software error occurred in MN but we are just ignoring it here. At least NJ is transparent & admits only a small fraction actually voted, which goes a long way to restore faith in the system.
English
1
0
0
8
Gregory Travis
Gregory Travis@greg_travis·
Saying that a system is vulnerable to contamination if it contains 0.0001% of a contaminate is either willfully lying or a demonstration of numeric illiteracy or both Every apple seed has a trace amount of cyanide in it. I don't go around telling people that apples are deadly to eat because the consequences of cyanide poisoning are dire So stop lying that elections are vulnerable to poisoning unless you're going to tell people to stop eating apples for the same reason
English
1
0
0
7
Ryan Lame
Ryan Lame@RyanLame1776·
@greg_travis @DonaldJTrumpJr I think you are missing the point. Election consequences are too high to have a vulnerable system. Personally I doubt people have been voting against their own self interests for years. Aka California and New York.
English
1
0
1
8

Keşfet

@SimonLermenAI @ramez @JeffLadish @GaryMarcus @mkratsios47 @KittyAnnabelle3 @ryangrim @RyanLame1776