In a normal world. this security analyst would be shitcanned and a new protocol for the use of AI in the DoD would be issued (obviously requiring independent vatting of any conclusions it makes…) If Trump is trying to sweep this under the rug, this would be spilling over into the “why the hell have both parties not already impeached this guy?” topic…
So I’m not the only one who sees this “dangerous” narrative is a scam, huh? The companies are dangerous, trying to evade their legal responsibilities by executing a shell game, not the tech.
Meidas Touch: Viral AI WARNING Just Got REALITY CHECK (link…)
Anthropic is desperately seeking to pospone their IPO. With DeepSeek in the wings, the IPO will eventually fall on its face. Thankfully all this is getting flagged PRE IPO! The AI monopolies are seeking to get subsidies in exchange for their postponing deployment. Fuck them! If their data centers drive up utility and real estate prices for voters, they need to be enjoined. If they are selling tokens for third parties to engage in reckless deployments (case in point, the intel misuse listed above…), they need to be sued to cinders by State AGs.
Just think about any of the cases many years before this where companies could have given tech to reckless third parties, and how the companies would be on the hook for that (case example: giving people copier tech that would vastly simplify currency counterfeiting; the companies making those high end copiers had to show the feds they were policing their shit, otherwise the secret service would put their asses in a sling…) The tech bro’s are childishly seeking to re-adjudicate the well-established legal precedents of implied contracts…
So is neglecting to secure the power at the time you started building the data center equivalent to an act of war or environmental catastrophe?
Or is it just bad decision making driven by greed and FOMO[1]??
If you’re a Fortune 1000 and not migrating off of Oracle already, you are making one hell of a BFM[2]…
The (Overdue) Collapse of Oracle’s $900B Empire (link…)
People are exiting to “postgreSQL”, which is an open source product -you can’t acquire that.
The fundamental problem with Ellison’s AI play is that he based it all on a company that is ethically challenged; OpenAI is not going anywhere with Apple suing them. He has one of two choices: sue OpenAI (and by the time that would play out, the money would not even be there! -it would only serve to accelerate the AI bubble bust…), or fall on his sword for having made a bad bet. His “force majeure” gambit is a childish JOKE to evade responsibility for a bad business decision.
I have seen this AI movie once before. In the dot com era, companies were marketing websites as a business model (often when they did not deliver a good or service -“well, in the future…”) I would get from people when I asked questions: “You just don’t get it…” Then the dot com collapse happened. Thank Dobbs I never “got it”!
I am getting that same “you just don’t get it” line from AI psychotics right now, in spite of all the glaring liabilities that AI makers will be on the hook for AND the models not being reliable!
I have to reprint my disclosure: AI can be good for code auditing, and thus improve software quality. But DeepSeek is undercutting the American monopolies on the price to accomplish that. There is NO WAY AI can do truly generative stuff; what they are calling “AI” is only an option when failure can be an option (and on occasion, it can be if you’re doing research or run-it-up-the-flagpoles -but not for most business use needs!)
We need state-by-state legislation that spells out the onus of liability on not only AI implementers (geniuses who are hooking up models to critical infrastructure -the air traffic controllers would be a good starting place for that!), but also for the paid model providers whom are enlisted. I would rather see a consistent federal enforcement, but given all the opinions in play, I think having a model of competition among States is more realistic.
States that over-regulate will end up with nothing getting done; those that under-regulate and end up with catastrophes on their hands can inherit the calumny of being reckless…
Who has been pointing out the liability issues?
Lil’ ol’ me??
You need State regs that put the burden of responsibility on the [paid] model providers and the agent implementers…
Congratulations, NVIDIA; now YOU will get sued if that doesn’t work out!
But I guess the alternative of watching the circular financing collapse was not a great prospect…
I’m shocked! Shocked!!
Breaking Points: Trump, CEOs Parade SHAM AI Deal To Self Police (link…)
STATE regulations is where it will be! Very hard for the AI Psychotics to stop them. Federal? Excuse me while it gets tied up for years -just look at the Dumbass -er, “Genius”- Act.
I am most pleased that prog pundits are catching on to the “AI is dangerous” distraction/deflection! It’s simply a matter of liability, both for the model makers and the agent implementers. The State regulations should put the burden of proof on the makers, especially when damage has been done. The core of the AI con (not that all AI is a con, but like crypto most is now…) is to reassign agency away from the companies. If you get harmed, they expect you to go to civil court and chase your tail…
If you have serious penalties in State regulations for recklessness, they will be sure to do such audits, regardless of what they say. None of this fucking “Trust me Bro’!” horseshit! (OR: “Someday we’ll have federal regulations!” -yeah, right!)
Some of the monopolies’ beancounters might reassess the profitability of the business model, and eventually pass on it.
The States could choose to regulate AI ventures one of two ways:
-
Require insurance for a proscribed capital amount. This already exists as a standard in the software industry: developers on high-profile projects are often required to carry “errors and omissions” insurance.
OR
-
Require AI ventures to be bonded (effectively self-insured…) The State would proscribe an amount to be earmarked. This is in line with crypto regulations that are materializing now: stablecoins have to be audited and have a balance on reserve for the amount they issue.
The idea is: if damage is done by them, injured parties don’t have to chase after them in civil court with expensive lawyers to recover for the damages. Worse yet: what happens if the business ventures just decide to Chapter 11 or 4 and walk away from responsibility for the damages??
What’s the best approach? Well, that’s the beauty of doing it at the State level: you can have a marketplace of methods. After a year or two, it will become evident which ones are practical and which ones are not. States that had few or no regulations and voters got harmed? The voters can place the heads of the State politicos who encouraged that on a pike. AI monopolies will grouse about having to follow different standards; but that’s some tough fucking shit! Those companies should not be so outsized to begin with.
The regulations should be for BOTH the AI model monopolies that charge for their access and the cottage industries that are wiring up agents to the models.
Now you can see why the Tech Bros’ were swooning over AI: they were hoping it would be an opportunity in the tech sphere to rewrite legal precedents on liability and implied contracts! “It wasn’t us who helped your bank account to get drained, you to get run over by a self-driving vehicle, or you to be harmed in any of a number of other technical schemes! It was the AI’s fault!!” It was a childish attempt to assign agency to their creations and pass the legal buck to inanimate products that they got paid for.
The point of State regulations for insurance and bonding is to force the Bros’ to have some skin in the game. They would argue they already put skin in the game through development. Puh-lease! You jokers who violated intellectual property law at every turn and who used open source at every opportunity but gave precious little back to open source?! What a bunch of entitled man-children!!
I see Newsom/California has just dipped their toes into AI regulation.
This will be a holy war for the Bros’.
NO ENFORCABLE LIABILITY
So we realize it at the State level!
Feel free to do the robot thing, but AI without liability controls? Nah-ah!
You tech bros’ are screwed; not us. I disagree with Cenk’s analysis that it screws all of us (link…) Thankfully, it was caught in time before it could have gotten much worse, and now we have to have our dot com phase of retrenchment, which we got through once before in tech. Yes: it’s a much bigger capital lockup than dot com was, but it has not been tied to retail investors yet. The VCs betting the farms, the AI monopoly players -THEY are the ones who get the shaft; which we will certainly feel, but we can survive that. The AI doomsday-if we don’t-do-what-they-say narrative is just another setup for a bailout heist.
That is fun-nee!
Reminds me of this Star Trek NG[1] episode…
Next Generation ↩︎
Forbes: AI Execs Asked ‘Raise Your Hand If Your Company Has Insurance For Catastrophic Risk’ (link…)
And THAT, ladies and gentlemen, is the fundamental issue!
They had nothing to say! How could they NOT have expected that as a question?! You would think they would have had an argument at the ready…
The Tech Lead: “There’s no such thing as Rogue AI” (link…)
This guy astutely points out how hackers got extradited for attacking systems -but no such consequences for the AI monopolies. The AI monopolies would argue: “Well, unlike those hackers, we didn’t intend to do break ins…” You had your software, which you were PAID for, do it because of YOUR reckless design that had little forethought as to safety controls.
Turns out, as I had said, “Mythos” is not the sexy sell they had claimed it was. Note the Department of Treasury’s (Bessent’s) complicity in hyping it -to boost the markets and larger economic nymbers?
Mythos: not the magical AI hacking tool after all (link…)
Dobbs bless tech pundits like this guy! Until about a year ago, “AI” was a grifting hype machine with no pushback whatsoever from commercial media. Now, the bill has come due for their fraudulent press kits and astroturf news websites with guys like this getting out the hard numbers…
China Is About To Pop The AI Bubble (link…)
China is already upping their token prices; the compute is not free! But they will always undercut the US AI monopolies. The tech bro’ dinner at the White House with Xi was a pathetic attempt to form a tech cartel and keep the prices up artificially. I’m sure Xi said “I’ll take it into consideration” (yeah, right…)
Patrie, I agree with you on some things, and disagree with you on some other things.
What I disagree with you on, is when you say AI sucks, and it is only for things that can fail, and it not for actual business use. I worked as a software engineer for almost 20 years and will probably never write another line of code myself ever again. From here on, everything I build will be from prompting AI’s. It doesn’t suck anymore. It’s good enough to use for production.
What I agree with you on is that these AI companies are over-blowing the risk that it poses to the computing infrastructure. At the end of the day, AI is just software. It can only do what software can do. Additionally, what software can’t do, AI can’t do. The hysteria surrounding AI is based on the general populations lack of knowledge of technology. The idea that “we can’t control AI, so we have to ban it” is just so absurd. Software is something we know well how to control. AI companies lying by saying it can’t be controlled is just so dumb. Unfortunately since TYT and all the other media companies have very few people that know anything about technology, they end up believing the lies.
-
Check out the ton of youtube videos now on how bad it is.
-
Take out your AI prompted code 2 years from now and make a 3% logic change. Have fun! Take out code you or someone else wrote well and has essential comments? Straightforward.
-
It’s not reliable; we have to agree to disagree on that point. If it’s reusing code that has been battle-tested for years, it will be reliable -but you don’t know that! And it can’t always reliably integrate reliable subcomponents.
I will certainly start to use AI to audit code, provided it can be done cheaply. But no way will I be the fall guy for the statistical result of having consulted the Internet for a solution.
A funny anecdote on AI involving Linus Torvalds (he has blown hot and cold on it; for example, he says desktop linux is a failure -beg to disagree, if you are using a version from the last couple of years…) His latest complaint is that many exploit reports are not being triaged, so their Inbox is getting jammed. Typo’s in diagnostic messages had been getting classified as exploits (!)
Linus was working on a bug in a video driver. He dialed up an AI and asked for assistance. What did it say? “You can’t fix that bug; it’s impossible to fix.” After working on it a lot more on his own, he got it fixed.
I know how that AI came to that statement. On StackOverflow you will see someone post a problem; then SOMTIMES someone comes along and patly states it’s not possibe to solve the problem. Without rubbing the nose of the poster in it, I then will sometimes say: “Actually there is a solution; you could do this…”
So the AI ingested that “not possible” viewpoint and spat it back out. See a problem here with the statistical backreference approach? It can only be as good as your training data -which often is quite bad!
If an AI says I committed a bug, I will always heed what it says and VERIFY that is correct. And if it’s not expensive, I’ll use it. But I will not allow it to fling out code. There are just too many cases documented online now of the crap it turned out as being unmaintainable, lurking with errors, and next to impossible to modify or have it modify (unless we are talking about cosmetic changes, which are often not what needs to be done…)