US judge orders lawyers to sign AI pledge, warning chatbots 'make stuff up'

I wonder if this had anything to do with it?
https://arstechnica.com/tech-policy...e-up-by-chatgpt-judge-calls-it-unprecedented/

 

The traditional job of Proofreader may be making a comeback. Unfortunately, in today's electronic communications world, there has been too much reliance on spell checkers which have no ability to correct poor word usage. What is even worse is speech to text converters which are horrible for correct word usage and substitute "Phonetic" variations that are horribly incorrect usages.

I've found that the grammar checkers that come with Windows are not too bad. But, like all of this stuff, you can't just accept what they offer blindly. You need to judge for yourself whether to take the recommendations, but I would guess that I agree with the recommendations about 2/3 of the time, possibly more. Also, I've learned a few things about grammar over the years from the explanations that are given.

 

I wonder if this had anything to do with it?
https://arstechnica.com/tech-policy...e-up-by-chatgpt-judge-calls-it-unprecedented/

"It's called a chat bot, not a research bot"

What would he do?

 

https://www.empireonline.com/tv/news/charlie-brooker-chatgpt-black-mirror-episode-exclusive-image/

“I’ve toyed around with ChatGPT a bit," Brooker reveals in the new issue of Empire. "The first thing I did was type 'generate Black Mirror episode' and it comes up with something that, at first glance, reads plausibly, but on second glance, is $h!t. Because all it’s done is look up all the synopses of Black Mirror episodes, and sort of mush them together. Then if you dig a bit more deeply you go, 'Oh, there’s not actually any real original thought here.' It’s [1970s impressionist] Mike Yarwood — there’s a topical reference.”

Artificial intelligence is actually really stupid. ChatGPT will never be creative.

 

https://www.empireonline.com/tv/news/charlie-brooker-chatgpt-black-mirror-episode-exclusive-image/


Artificial intelligence is actually really stupid. ChatGPT will never be creative.

I don't know that I would go that far. Look at how many shows and movies and books have come out in the last (pick your number) of decades. What fraction of them can truly be described as "creative" or "original"? Almost everything that humans produce in these fields is about as creative or original as the tripe ChatGPT and its ilk are spewing forth. So why should the standard for ChatGPT-like "authors" be any higher?

While I might expect a "typical" human-generated story to hang together better than an AI-generated one, it's only because the human is specifically trying to make connections between different parts of the story, while the AI is just stringing words together based on probabilities. But even with human writers intentionally trying to make plots hang together, they typically do a very piss-poor job of it, leaving gaping inconsistencies and holes in the work. But audiences, by and large, just ignore that, provided there's enough sex and explosions and technobabble. The difference with AI-generated stuff, is that we are not ignoring those inconsistencies and holes.

Here would be an interesting experiment: What fraction of stories that are pitched to a movie company or a TV show or a publisher actually get accepted? I have no idea, but let's call it X for some particular genre. Have ChatGPT produce a bunch of episodes of some TV show, for example, and pick a fraction of them equal to X (not randomly -- use the same process that is used for human-generated stories). Then actually produce some of them and put them in with all the rest and let audiences (even if just test screenings) view a mix of them without knowing that some of them are AI-generated and see how they rank. While it would be nice to think that audiences would immediately be put off by the AI-generated tripe, the fact that they are seldom put off by the human-generated tripe makes me dubious that this would be the result.

Here's another idea that might actually be not only useful, but also a worthwhile achievement in the development of AI. Develop an AI with the ability to review a manuscript and point out inconsistencies, holes in reasoning, and things that defy physics. That seems like a real challenge, but if achieved, could be used as a real tool to help improve stories used for movies and books and TV shows.

Having said that, I don't believe these industries care. They already have technical advisors (I've known several) who routinely point out the glaring crap in the script and who are almost always ignored. Their sole role is not to have any actual impact on the product, but merely to be window dressing to be held up and pointed to when it's convenient to comment about how they "have a team of highly-qualified technical advisors who review every part of the script for accuracy and authenticity." Sure, that might be a true statement. What's missing is the fact that their reviews fall of deaf ears. The producers don't care. They have a very good understanding of not only what audiences will tolerate, but what audiences want and expect and will pay for. Accuracy is near the bottom of the list, while superficial plausibility, even if it can't stand up to the slightest bit of examination, is near the top.

 

In the 80s AI was being bandied about and there were startups trying to make money off of AI software mostly targeting Field Service Engineers to improve diagnostics and repair of various equipment. One startup company mentioned in a trade journal had a major customer that they were dealing with and trying to get better performance numbers out of their AI software. They had gone beyond 50% accuracy, moved beyond 60% and were still pushing to meet the customer's request of 80% plus accuracy. The kicker of the story which was reported in the trade journal was that the customer was a major Las Vegas Casinos sports betting bookie trying to use AI software to set the line on professional football and baseball games.

 

If most humans in the field are not creative or original, the tripe (at least tripe is edible ) from ChatGPT that used that human writing as input, has no hope of even being more than a mediocre blend of leftovers and spoiled beer.

 

If most humans in the field are not creative or original, the tripe (at least tripe is edible ) from ChatGPT that used that human writing as input, has no hope of even being more than a mediocre blend of leftovers and spoiled beer.

Agree. And this is the point that all of the ChatGPT fans seem unable to grasp. Garbage in, Garbage out. If the AI training data is thoroughly polluted with inaccurate content, how can it be reasonable to expect the output that results to not be? But these people tend to come from a "big data" mindset that believes that only the "three V's" matter when it comes to large data sets -- volume, velocity, and variety, They don't care about validity or veracity and claim that, like working with large sample sizes in signal processing, the noise will average out leaving only the true signal. I think that, in many cases (including these large language models) they are making a couple of faulty assumptions. First, they are underestimating (if they ever actually did an estimate in the first place) the low signal-to-noise ratio of the content on the Internet, and second, they are assuming that the noise is random such that it tends to cancel out other noisy content and don't realize the degree to which the noise is not only systemic, but self-reinforcing.

 

I wonder what tests one could use to tell if something produced was from a AI engine, a Human, some other animal ?
whats replaced the old Turing test ?

 

I wonder what tests one could use to tell if something produced was from a AI engine, a Human, some other animal ?
whats replaced the old Turing test ?

Nothing, it was a test created by a brilliant mind that greatly underestimated what a mind was..
https://en.wikipedia.org/wiki/Chinese_room

Searle's thought experiment begins with this hypothetical premise: suppose that artificial intelligence research has succeeded in constructing a computer that behaves as if it understands Chinese. It takes Chinese characters as input and, by following the instructions of a computer program, produces other Chinese characters, which it presents as output. Suppose, says Searle, that this computer performs its task so convincingly that it comfortably passes the Turing test: it convinces a human Chinese speaker that the program is itself a live Chinese speaker. To all of the questions that the person asks, it makes appropriate responses, such that any Chinese speaker would be convinced that they are talking to another Chinese-speaking human being.

The question Searle wants to answer is this: does the machine literally "understand" Chinese? Or is it merely simulating the ability to understand Chinese?[6][c] Searle calls the first position "strong AI" and the latter "weak AI".[d]

Searle does not disagree with the notion that machines can have consciousness and understanding, because, as he writes, "we are precisely such machines".[5] Searle holds that the brain is, in fact, a machine, but that the brain gives rise to consciousness and understanding using machinery that is non-computational. If neuroscience is able to isolate the mechanical process that gives rise to consciousness, then Searle grants that it may be possible to create machines that have consciousness and understanding. However, without the specific machinery required, Searle does not believe that consciousness can occur.

 

Nothing, it was a test created by a brilliant mind that greatly underestimated what a mind was..
https://en.wikipedia.org/wiki/Chinese_room

And at this point, our brains are to ourselves nothing more than mere black boxes. That is, information goes in and information comes out. And we know nothing about the process that takes place inside said boxes. Except that we know and we are aware that we are said boxes.

In fact, each of us has to make a leap of faith of sorts and trust our senses and our reasoning into accepting that our peers have a mind and a consciousness not unlike our own. Even though we still have no way of proving that other consciousness but our own actually exists.

How then, I wonder, will we ever be able to prove that an AI has reached the singularity of consciousness?

 

Even though we still have no way of proving that other consciousness but our own actually exists.

Yet most will trust another's mind before they trust their own.

 

Yet most will trust another's mind before they trust their own.

I'd say that we have to ... as a species, we have to trust Newton, Einstein et all if we want to move forward and not try to start from scratch every time we want to learn something.

Practically all of our human knowledge is inherited. There's only a small pinch that each generation contributes to the next one. And yet the cumulative effects have been astronomical in the last few decades.

Plenty of positive things to be grateful for nowadays, I'd say. Science and technology have made our lives the hell of a lot easier than our grandparents. I'm an optimistic at heart in that respect.

 

Yet most will trust another's mind before they trust their own.

Most of us know, the most easily fooled mind by our actions, is our own mind.

 

I wonder what tests one could use to tell if something produced was from a AI engine, a Human, some other animal ?
whats replaced the old Turing test ?

I don't think there's anything that needs replacing yet. Remember, a machine does not pass the Turing test by having a single instance of a single person interacting with it and being fooled. It passes the Turing test if there is no reliable way for any evaluator to observe the interactions between a human and the computer (without knowing which is which) and be able to determine which was which. The evaluator doesn't have to be right every time. Imagine that thousands of evaluators each observes thousands of sessions. The Turing Test would be considered failed if, collectively, the evaluators were able to make the determination correctly in a statistically significant fraction of instances.

One subtle point that people that pull out the Turing Test often forget is that the Turing Test is explicitly NOT concerned with the accuracy or correctness of what it stated, only on whether it gives answers that were statistically indistinguishable from the answers humans would give. In this sense, ChatGPT and the like could conceivably get to a point where they could pass the Turing Test even if their ability to produce accurate results didn't progress much further than it is right now -- anyone whose ever graded essay questions can attest to that.

Searle's Chinese Room looks at things from a different perspective that tries to make the distinction not on whether a machine could successfully mimic human interaction, but whether it can actually understand what it is producing, claiming that if it can achieve the mimic behavior without understanding, that it is simply following a sufficiently sophisticated algorithm. But even here there are problems -- humans do all kinds of things where they are just "following a recipe" with little or no understanding of what they are doing. The key point, however, is that they don't do this for everything, and the capacity to do it at all is the distinction.

 

I'd say that we have to ... as a species, we have to trust Newton, Einstein et all if we want to move forward and not try to start from scratch every time we want to learn something.

Practically all of our human knowledge is inherited. There's only a small pinch that each generation contributes to the next one. And yet the cumulative effects have been astronomical in the last few decades.

Plenty of positive things to be grateful for nowadays, I'd say. Science and technology have made our lives the hell of a lot easier than our grandparents. I'm an optimistic at heart in that respect.

Traditional science does not rely on trust. Only this new-agey post-modern crap they try to convince us is science requires us to abdicate our own minds.

Never trust a scientist who says, "trust me."

 

Never trust a scientist who says, "trust me."

Yeah... trust should be applied to politicians only.

 

Yeah... trust should be applied to politicians only.

More and more there's no difference.

The Age of Reason is past. We are in a new dark age.

 

Traditional science does not rely on trust. Only this new-agey post-modern crap they try to convince us is science requires us to abdicate our own minds.

Never trust a scientist who says, "trust me."

I have never seen anyone who is a “scientist” say “trust me” concerning a topic of scientific investigation. What makes a person a scientist is the practice of science which is a combination of the scientific method constantly seeking new and better ways to confirm theories.

Wearing a lab coat doesn’t make someone a “scientist”, and a scientist that merits the title is never more of an authority than they can show by evidence and logic. Popular reporting of science and scientists is also not science.

Skepticism of all human endeavors is surely warranted. Even real scientists are also human and suffer from the shortcomings of the human relationship with "the truth”. But this is precisely what science as a practice is intended to combat. Systematic, socially driven errors in the conclusions of scientists as a group can and do certainly exist but it can’t, isn’t, and will never be because that say “trust me”.

It is easy to be jaded about people and society and translate that into “wise” pronouncements on things we aren’t actually equipped to understand, but this isn’t wisdom it’s just compensation. Genuine skepticism seeks answers, it’s not just a denial of anything that doesn’t serve to advance our own agendas.

[edited for typos]

 

You are a relic. And naive. Our children are not taught to think this way anymore -- or at all. They are taught to believe and obey, or be destroyed.

I have no idea why you imagine you have access to what it “really” going on, but you don’t. You are not more informed because you only look for conspiracy and political agenda—you are actually less so because your confirmation bias (something we all have) gets used to prove (to yourself) that your political beliefs are correct.

You play the skeptic but you are only a skeptic about others, not your own opinions. I have seen from your posts that you think too highly of how well you understand the technicalities of many things you comment on. I‘ve simply ignored it because I have little hope of influencing you at all. I guess I got tired of it and decided to say something about this one.

You get to have the last word because despite the earlier wisdom of ignoring your posts I made the silly error of responding to one. I won’t make that mistake again.