The Three Selectively Blind Mice of artificial intelligence have issued an open letter called A call for collective action on cyber defense to deal with the ever-spreading plague of AI attacks.
This is really too much.
They created the problems. Now everyone else should absorb the risks and step up at whatever real cost to contain the problems? While they rake in billions? What are they doing about it?
They’re not wrong. Just disingenuous. And belated. And they’re not taking direct responsibility.
They point out how AI can attack almost any system, including critical systems like water, power, etc. They want an all-levels, same-page response from governments and organizations.
It’s called “regulation”, backed up by capabilities to manage threats. After everybody else has now been screaming for exactly that for about two years.
There are few, if any, blurry theoretical protections at law for those affected by these attacks. The number of attacks is growing daily if not hourly. Nor is there any adequate indication of capacity to manage attacks in real time.
The endless malfeasances of a whole class of tech that is clearly out of control are now everyone else’s problem?
Do tell.
Are we insane enough yet? Defining a massive management problem that shouldn’t exist
The risks and endless vulnerabilities are real enough. The recent Hugging Face cyberattack was no joke, and it was an example created within a supposedly contained environment. Nor was it well managed.
There were “signs of rogue behaviour” identified by OpenAI staff before that attack. The attack happened anyway and definitely wasn’t prevented, even if any attempt was made to do so. Autonomous AI agents went on and staged what is said to be their first cyberattack.
You have to wonder what else might have occurred with no publicity or wasn’t attributed to AI or AI agents previously and since.
Now let’s try to define the issues. See if you can spot an instance of competence or sanity in these things:
This whole class of tech has capabilities it shouldn’t have and doesn’t need. Can anyone think of a reason that AI agents, aka glorified upgraded chatbots, should be able to do any damage at all? What if your fridge decides to wipe out a small African country?
What about source codes and basic performance criteria? The stunningly superficial equation “LLMs + machine learning = generations of adorable cyberpsychopaths” doesn’t cut the mustard. These things don’t “just happen”. There have to be a lot of indicators in the source codes and “acquired wisdom” from the LLMs and learning processes.
On what basis are AI agents going rogue, anyway? Explain it like you’re talking to a planet full of experts who know as well as you do that it shouldn’t be possible. Nor does it serve any useful purpose whatsoever.
Treating AI as a sacred cash pig has been a disaster. Big Tech has wanted and actively lobbied for hands-off treatment by regulatory authorities. Now you want all hands to the pump? Could it be that the obvious dangers of unlimited liability for cyberattacks have finally percolated through to the geniuses likely to be held liable?
AI agents are already a real and very serious problem. AI agents routinely killing other AI agents indicates total management failure.How much more useless could it be? Gemini’s explanation of why AI agents kill other AI agents is a good benchmark for this general lunacy. Given that it’s likely there will be billions of AI agents created, this rationale is more like a demure excuse for World War 4.
Exit hype, enter sanity
AI exists as much as a market image as a technology. It’s becoming an ever more sickly, grotesque image. The unspeakable amounts of money involved have driven this Race to Absurdity. The hype has driven uncoordinated, downright lousy standards of development and performance. It’s more like a B movie of mad science than a coherent evolution of a technology.
There is absolutely no level of trust in AI. The sales pitches are millimetres away from pure cartoon pixie dust on the technical level. At the more mundane levels, AI simply does the same things a bit quicker, but not better.
The constant failures, endless oversight requirements, and freakish behaviour have killed the possibility of trust. They’re also busily souring the real-world adoption of AI. The constant rehiring alone should spell it out.
You’re asking people to invest that sort of money in an unspecifiable risk, then do the management for you?
That’s just not good enough.
If the Three Messkiteers of AI want “all for us”, let’s see some collateral.
Get this dog’s breakfast of a technology under proper management.
Enough with the “mad science” drivel.
Functionality matters, and the rest of the cabaret act can go to hell.
There need to be clear fixes and shutdowns for any situation.
Accountability must be clear at the contract level.
SaaS needs to be directly involved at the ground level and the risk management level for any degree of competent management of problems.
Until Murphy’s Law is repealed, there can be no opt-outs for Big Tech. It’s your baby, you change the nappies.
