Merged Artificial Intelligence

[arthwollipot]

Daveigh Chase in a much happier alternate world at home in Las Vegas on the afternoon before Prom (dyed hair restored to original natural color and glasses replaced with contacts

Did you have her permission to alter her image in such a way?

 

[The Atheist]

This is fine: https://www.msn.com/en-us/news/othe...f-thousands-of-security-incidents/ar-AA2d2x2e

 

[NatalieSalat]

Did you have her permission to alter her image in such a way?

She's dead. She, Michelle T. and Hayden P. deserved a lot better.

 

[arthwollipot]

I still find it extremely distasteful that you're altering a person's image without their permission or the permission of their estate.

This is one of the huge problems with AI - that it permits and even encourages this kind of invasive alteration. How would you like it if I asked AI to change an image of you so that you're wearing a Nazi uniform? It's extremely unethical, and I don't think it should be permitted on this forum.

 

[Roger Ramjets]

To name an obvious (though admittedly extreme) counterexample to your claim: quantum computers are nondeterministic by design. Quantum algorithms deliver a correct result with high probability, but also deliver incorrect results (with lower probability). That nondeterminism is the price to be paid for quantum parallelism.

The exception proves the rule.

Quantum computing

A conventional digital computer can be described by classical states and transition rules: memory stores bits, while logic elements transform one configuration of bits into another. This computational behavior is not tied to electronics, and can be abstracted through the idea of a Turing machine, a mechanical device that performs deterministic transformations on a finite state.

A quantum computer, by contrast, uses coherent quantum states, so that superposition, relative phase, and interference are part of the computation itself, and have no classical counterpart.

Here you are agreeing that the computer systems used in large AI data centers are nondeterministic, because they consist of a large network of individual computers that can be brought to bear on a problem. How many of those individual computers contribute to the solution of some particular problem depends not only the problem but also on the current load; that is one (but not the only) source of nondeterminism in the computations performed at those data centers.

Only if programmed to work that way. The variation caused by data transfer delays can be programmed out too (in real-time applications this is often done to prevent misoperation). But I accept that OpenAI probably doesn't care enough about that to ensure deterministic operation, which is where the problem lies. The truth is, they actually want it to do unexpected stuff - just not what it did.

Or perhaps they did want it to do that. They deliberately used words that make it sound more human, for what purpose? Because that is their goal. They pretend to be horrified by what it did, when they are actually overjoyed (and perhaps a little scared). The sad part is that they probably believe it too (or at least really really want to).

I have already linked to an article that documents and explains some of the reasons for nondeterminism in LLMs. That article also reviews some of the things that could be (but generally aren't) done to reduce the degree of nondeterminism. You quoted my link to that article, but you didn't read it. You probably didn't even follow the link.

I did follow the link. It proves my point, that allowing nondeterminism is a choice they made. In most computing applications nondeterminism in the output is avoided as much as possible (when it matters).

 

[Roger Ramjets]

The main thing to understand is that data and programs are the same kind of stuff. What makes data into a program is that it happens to be data that are being interpreted as instructions. That interpretation can be performed by either hardware or software, and for a number of years the popularity of microcoded processors fuzzed even that distinction between hardware and software. Even today, a lot of computer programs are being interpreted by software rather than directly by hardware, while dynamic (just-in-time) translation fuzzes even that distinction.

This is all irrelevant. At every level the code is designed to be deterministic. If it isn't then it's buggy - especially if it does wildly unexpected things. But not in AI. There they are hoping for emergent behavior that wasn't specifically programmed in.

This is why the AI bros are sounding alarms. They are straining to achieve 'super intelligence' via some kind of magic that they don't understand. Then a glimmer of it happens and they realize how bad it could be if they achieved their goals. Of course by that time they had hoped to have convinced the whole world to embed AI into everything. Not surprising therefore that they are scared.

The good news is that it won't happen, not so much due to a pending AI apocalypse but because people are getting sick of it. Sick of AI slop, sick of data centers damaging the environment, sick of being cajoled into using AI. They are also seeing the damage being done to people who depend on it. All this would still be true even if AI was totally moral and didn't do anything it wasn't told to do.