Oh no, they told it to "go rogue" and it went rogue. We better ask the government to regulate our competitors while we secretly continue developing this dangerous technology for the select few with insider connection. You know, for national security reasons. Now the AI is aligned for the good of humanity. You can trust us because we're the good guys, move along.
Computers are used to evaluate LLMs, but LLMs are not "software" or "algorithms" in the traditional sense. They are not built out of conditional branches or loops.
So trying to squeeze the observed behavior of this new thing under existing terms like "software bug" is at least as much of a force-fit, and what you're doing here is just as much language engineering as choosing to use a term like '[mis]alignment'. Which is fine, this is just one way that humans choose language.
LLMs are vectorial databases with losses that index statistically filled data, which uses a text interface to query such statistically filled data. The output is a string concatenation.
By the nature of the used architecture in such software, the used algorithms, when queried (prompted), you can get random mixed data as output, ERRORS, due to undesired indexes getting closer at one point while the string was being concatenated for the output, what affects the rest of the indexed content that will be concatenated.
And this is intrinsic to this tech. The larger the context, the greater the probability of get mixed data. And if the provider lowers the precision of those indexes -in order to decrease hardware and energy resources consumption- such probability increases to the point where those errors are granted.
Anyway, even knowing that the queries can return wrong/mixed data in the responses (errors), the companies developing this, decided to introduce a new product, that connects such LLMs responses to the command console, latter connected to internet, running commands from such returned responses witch obviously can contain whatever mixed random. Then we started to hear "oh, it deleted my directory", etc.
Again, One have such described statistical database with text interface, witch query the database recursively with the output text of the previous query, and this is connected to the command console. Larger contexts, several times... What should we expect as result? rhetoric question.
Implying sentience or consciousness is a convenient marketing strategy that has been introduced by anthropomorphising the names of all the methods and algorithms used. An "Agent" should be translated from such deceiving language to "context splitter querying in loop that consumes more tokens from us", or similar.
* > An "Agent" should be translated from such deceiving language to "context splitter querying in loop that consumes more tokens from us", or similar.
Please disregard this line. I wanted to point out that it increases the length of the context (and therefore the probability of errors) due the batch processing. But I redacted it incorrectly because I also wanted to imply that promoting the use such queries non-stop increases the billing through tokens consumption.
Not being built out of conditional branches or loops does not mean they’re somehow outside algorithms or computation. Learned parameters don’t confer exemption from computing.
Did the engineered system behave as intended? No? Then you’ve got a gd bug/failure.
LLMs run on computers and are thus constrained by the capacity of that which runs it. If the system running the LLM has no network and no software or software-tooling, how does the LLM's generated text take action on a system(computer) that requires software to do anything?
Also, I absolutely agree LLMs are not software, and thats my point. LLMs without supporting software tooling surrounding it cannot do anything but print text. And even the printing of that text happens through software
The answer is: It's irrelevant, because no one runs LLMs on systems without networks or missiles or some other way to "take action" because that would be pointless.
So you agree, its the computer components that actually do the thing that is important, thus we should be talking about those components (software-tooling) which do the things
I think it probably has more to do with Mac's lack of adoption by the US Government and large corporations.
Microsoft took accessibility of Windows and apps as a basic requirement. The fact that Windows is the daily driver desktop for the staff of the regulatory agencies that would bring complaints would not have gone unnoticed.
But really, why wouldn't you? You almost get it for free if you use the standard widgets, and smooth keyboard navigation is far faster for experienced users than having to repeatedly move one's hand between keyboard and mouse.
The scam here is presenting average occupational accident rates as a measure of "safety" when the obvious risk of nuclear power is the rare large-scale events which render large areas of land uninhabitable for generations while producing few directly-attributable deaths.
Are those large-scale events in the room with us? Or are you saying over the last, however many decades of all sorts of countries running nuclear plants, we've just been oh so lucky to have a negligible number of serious accidents? Luck is an interesting argument.
Also, you may want to fact-check your "large areas of land" because you're just making that up, if talking about accidents that have already happened, or going of by feels I guess if talking about hypothetical accidents that have by luck not happened all this time?
The real scam is bikeshedding about the safety of nuclear vs solar while happily burning coal. Take a deep breath, let your lungs taste the byproducts of burning coal. And now, I dare you, say something about the nuclear waste "problem". barf
Data centers are only 5-10% of all US electric power demand. A significant fraction of that load is responsive to spot market price. DCs incorporate their own battery storage infrastructure for backup purposes and some even sell stored energy back to the grid during times of high demand.
The technology was given almost a century to prove itself cost-effective. It did not succeed in the market and only ever worked with truly massive amounts of government subsidies.
High fixed costs + more reactor sites -> more expensive, more nuclear waste sites for taxpayers to clean up
Each of those existing plants were competitive against coal at the time without having the negative externalities of widespread radiation exposure (uranium in coal just released to the atmosphere); mercury contamination in all our freshwater lakes and the ocean, leading to strict limits on consuming fish for children and pregnant women; and a half-century of CO2 emissions.
Orders of magnitude more people are killed by rooftop solar, but we haven't raised safety standards on all other sources of electricity to be the same level we require for nuclear.
Oh good, they told it not to "go rogue" and it just stopped doing it.
Nothing to see here, the AI is aligned now, move along.
reply