https://www.creativebloq.com/ai/ai-...on-winner-still-hopes-to-copyright-his-ai-art
People are "blatantly stealing my work," AI artist complains

Now, he's launched an appeal against that decision, claiming that his "624 iterations" with Midjourney, required at least 110 hours of human work. In an argument that many artists will surely find ironic, he also claims that unauthorised use of his artwork has resulted in him losing “several million dollars”.

“The Copyright Office’s refusal to register Theatre D’Opera Spatial has put me in a terrible position, with no recourse against others who are blatantly and repeatedly stealing my work without compensation or credit.”

While I'm sure there are all kinds of subtle, but legitimate, nuances involved, the basic guiding premise would seem to be pretty straight-forward.

Using music as an example, but applicable to most things, if I go and watch another artist and study their work and then produce "my" music such that it sounds just like theirs with perhaps some superficial tweaks, such as changing a few words or modifying a few chords, then I am almost certainly in violation of that artist's copyright. But if their work merely serves as one of many influences on what I create and my music has some fleeting similarities here and there to that artist's work, then I (in theory) have not violated that artist's copyright (contrary to what some artists think and demand and, occasionally, talk a jury into buying).

Whatever the rule was before generative AI, that rule should simply apply to work produced with generative AI. For an artist to insist that the use of their work to train a generative-AI engine because that can result in work that is similar to theirs, is very analogous to saying that no one that is ever going to create music can ever listen to that artist's work, since that can result in that person creating work that is similar to theirs. The violation isn't in listening to (or, for the generative AI engine, learning from) the work (provided it is accessed legally), the violation is in how it is used. If the generated work violates copyright, then it violates copyright regardless of how it was produced. Conversely, if the generated work does not violate copyright, then it does not matter how it was produced.

Now, what generative-AI engines greatly increase are the chances of someone creating work that is too similar to copyrighted work without intent to do so or knowledge that they have done so. But these are not new situations, either, so we should already know how to deal with them in our existing legal framework. What might need revisiting are how we deal with them because of the changing scale of the problem.

 

‘In awe’: scientists impressed by latest ChatGPT model o1

​

OpenAI says that its o1 series marks a step change in the company’s approach. The distinguishing feature of this artificial intelligence (AI) model, observers say, is that it has spent more time in certain stages of learning, and ‘thinks’ about its answers for longer, making it slower, but more capable — especially in areas in which right and wrong answers can be clearly defined. The firm adds that o1 “can reason through complex tasks and solve harder problems than previous models in science, coding, and math”. For now, o1-preview and o1-mini — a smaller, more cost-effective version suited to coding — are available to paying customers and certain developers on a trial basis. The company hasn’t released details about how many parameters or how much computing power lie behind the o1 models.

 

 

It mainly states the obvious.

 

It mainly states the obvious.

The very, very, very obvious.

The real takeaway, in so many of these stories, is not anything revealing about the technology, but much more revealing about the resilience of humans to insist on seeing things that simply aren't there.

 

The very, very, very obvious.

The real takeaway, in so many of these stories, is not anything revealing about the technology, but much more revealing about the resilience of humans to insist on seeing things that simply aren't there.

​

 

View attachment 333632​

https://www.wsj.com/tech/ai/yann-lecun-ai-meta-aa59e2f5
This AI Pioneer Thinks AI Is Dumber Than a Cat
Yann LeCun, an NYU professor and senior researcher at Meta Platforms, says warnings about the technology’s existential peril are ‘complete B.S.’

At the same time, he is convinced that today’s AIs aren’t, in any meaningful sense, intelligent—and that many others in the field, especially at AI startups, are ready to extrapolate its recent development in ways that he finds ridiculous.

Today’s models are really just predicting the next word in a text, he says. But they’re so good at this that they fool us. And because of their enormous memory capacity, they can seem to be reasoning, when in fact they’re merely regurgitating information they’ve already been trained on.
...
“We are used to the idea that people or entities that can express themselves, or manipulate language, are smart—but that’s not true,” says LeCun. “You can manipulate language and not be smart, and that’s basically what LLMs are demonstrating.”

 

Hello again,

I realized that probably the biggest problem with 'ai' will be that the designs will be done by companies that believe they know what 'ai' should be, rather than what it really should be. Right now, I see it as 'aai', which is artificial artificial intelligence ('a^2i'). In other words, the artificial intelligence right now is artificial in and of itself, in that real artificial intelligence would be actual intelligence of some kind, and the word 'artificial' seems to give the makers a sort of artistic license to do whatever they feel like doing then calling it 'ai'.

In even some other words, it should be "created intelligence", where it is actually intelligent. Not some ridiculous excuse for something that appears to be able to understand human language and provide some run of the mill answers that are probably not even correct.

My biggest fear, and I think I've mentioned this before, is that someone someday will read an 'ai' response and take it as fact, then use that fact for something important. That could be a disaster.

 

I searched for “internal resistance of lead acid batteries?”

AI reply:

The internal resistance of a lead-acid battery is usually very low, typically measured in milliohms (mΩ). A new flooded lead-acid battery may have an internal resistance of 10–15%, while a new AGM battery may have an internal resistance as low as 2%.

 

I searched for “internal resistance of lead acid batteries?”

AI reply:

Your typical AI mixture of good food and dog-crap that's polluting the knowledge base.

You can see this repeated now as if it were gospel.

https://www.everexceed.com/blog/answers-to-the-characteristics-of-lead-acid-batteries_b384

5. Internal Resistance and Power Output

The AGM battery’s internal resistance is among the lowest of the various lead acid batteries. While a new flooded lead acid battery can have an internal resistance of 10-15%, a new AGM battery can be as low as 2%.

https://www.autonationmobileservice.com/i/blog/agm-vs-lead-acid/
https://www.carparts.com/blog/agm-vs-standard-battery-which-is-better-for-your-car/
https://living.acg.aaa.com/auto/agm-battery-vs-lead-acid-battery
https://www.rbbattery.com/agm-vs-standard-battery-which-is-better/

All of the similar answers seem to come from the AI/ChatGPT era starting after some time in 2022. This is how synthetic data corrupts the datasets for future AI systems. The user generated patterns from AI searches start to reinforce the original AI mishmash into something that becomes the next likely words for a question instead of much better, more accurate, less similar responses from the past.

 

This is a problem that predated these LLMs. For many years it has been extremely common to find the exact same content on multiple websites, up to and including the exact same grammar, spelling, and punctuation errors, without any attribution whatsoever. One site simply grabs something wholesale from another site and then some other site grabs the same thing from one of them and the ball starts rolling. This was clearly being done automatically without regard to whether the content actually even matched what it claimed to be, so no attempt was being made to check for the accuracy of what was being plagiarized.

So it didn't take any sort of crystal ball to predict that the exact same thing would happen with LLM-created crap and that it would then start reenforcing itself in what would become nothing more than a huge electronic echo chamber.

 

I searched for “internal resistance of lead acid batteries?”

AI reply:

Yeah the responses seem to get worse instead of better. Very strange.
Yesterday I saw another bot reply that used percent for something that has actual units. Makes no sense at all.
This crap is just making the internet even stupider than it already was. How do we stop this junk.

 

This crap is just making the internet even stupider than it already was. How do we stop this junk.

We don't (or can't). Civilization is a self-defeating concept.

 

We don't (or can't). Civilization is a self-defeating concept.

That seems to be what I am starting to think also. We cannot see into the future so we can't know all of the true ramifications of what we think is an improvement that we make. The side issues that come up seem to dominate over the main issue that seems at first to be the improvement.

 

We cannot see into the future so we can't know all of the true ramifications of what we think is an improvement that we make.

You need to read more dystopian literature.

 

I searched for “internal resistance of lead acid batteries?”

AI reply:

So, checking the latest models for ChatGPT, I get:

The internal resistance of a lead-acid battery can vary depending on its size, type, and state of charge. Generally, it ranges from about 0.001 ohms for a large, fully charged automotive battery to around 0.02 ohms or higher for a smaller or older battery. These values can change as the battery ages or its condition changes.

The internal resistance of a lead-acid battery varies depending on several factors, including its size, age, temperature, and state of charge. Generally, for a fully charged, healthy lead-acid battery:

• Automotive (12V car battery): The internal resistance is typically between 2 to 5 milliohms (0.002 to 0.005 ohms).
• Deep-cycle batteries: These may have a slightly higher internal resistance, often ranging from 5 to 15 milliohms.
• Smaller lead-acid batteries (like those used in uninterruptible power supplies or emergency lighting): The internal resistance can be higher, sometimes up to 50 milliohms or more.

Factors affecting internal resistance:
• State of Charge: A discharged battery has higher internal resistance than a fully charged one.
• Temperature: Lower temperatures increase internal resistance, while higher temperatures decrease it.
• Age and Condition: As a battery ages or if it becomes sulfated or damaged, its internal resistance increases.
• Battery Design: The thickness of the plates, the type of electrolyte, and overall construction can influence internal resistance.

Why Internal Resistance Matters:
• Performance: Lower internal resistance allows the battery to deliver higher currents, which is crucial for applications like starting engines.
• Efficiency: High internal resistance leads to greater voltage drops under load, reducing the battery’s effective capacity.
• Heat Generation: Batteries with high internal resistance can generate excessive heat during charging and discharging, potentially shortening their lifespan.

For precise measurements, it’s best to consult the manufacturer’s specifications or measure the internal resistance using appropriate equipment like an impedance meter or specialized battery tester.

It is my experience that the models are improving and the ability to do realtime searches, as will as write and execute Python programs and use their output, has dramatically improved the quality of the responses. The addition of introspection and intentional fact checking in the 1o model has also had a big impact—though at a cost, since it uses much more compute.

I also note that the GIGO principle is very much at work here and the quality of the question has a direct effect on the quality of the answer. The models are more of a hypersophisticated search engine plus, than an oracle that can answer any question independent of user interaction shaping the response into contextually useful and accurate information.

I have found ChatGPT to be a very useful tool—one I am willing to pay for—in doing research and reasoning out complex logical issues interactively. In particular, once I have something refined, it is automatically documented which is very helpful. Of course I am not asking for factoids, and I am putting in some work.

 

So, checking the latest models for ChatGPT, I get:





It is my experience that the models are improving and the ability to do realtime searches, as will as write and execute Python programs and use their output, has dramatically improved the quality of the responses. The addition of introspection and intentional fact checking in the 1o model has also had a big impact—though at a cost, since it uses much more compute.

I also note that the GIGO principle is very much at work here and the quality of the question has a direct effect on the quality of the answer. The models are more of a hypersophisticated search engine plus, than an oracle that can answer any question independent of user interaction shaping the response into contextually useful and accurate information.

I have found ChatGPT to be a very useful tool—one I am willing to pay for—in doing research and reasoning out complex logical issues interactively. In particular, once I have something refined, it is automatically documented which is very helpful. Of course I am not asking for factoids, and I am putting in some work.

The problem is that it now starts takes experts (the actual oracle) to know when the result is a GIGO factoid search response because it looks so reasonable and complete at first glance even if it's a hallucination.

 

The problem is that it now starts takes experts (the actual oracle) to know when the result is a GIGO factoid search response because it looks so reasonable and complete at first glance even if it's a hallucination.

There is no doubt there is a large region of, possibly even most of. the apparent application domain for LLM ChatBots that has adverse outcomes. The motivation to profit and the profound ignorance of most people makes this a likely outcome, which is unfortunate because it can also be a damned useful tool.

On a related note, though it is only based on two datapoints I have a nagging feeling that the reliance on intensive computation, and so extreme power utilization, that characterizes both LLMs and cryptocurrency, points to something fundamental about a kind of technology that is not compatible with the society needed to create it and not available in a smaller environment.

 

ChatGPT is great for suggesting parts when you are designing a circuit.

Better than wading through the Digikey search engine for sure.

 

https://www.semafor.com/article/10/...a-centers-will-soon-hit-limits-of-energy-grid
Microsoft Azure CTO: US data centers will soon hit size limits

But as Microsoft and its rivals compete to build the world’s most powerful AI models, several factors, including America’s aging energy grid, will create a de facto cap on the size of a single data center, which soon could consume multiple gigawatts of power, equivalent to hundreds of thousands of homes.

https://www.wsj.com/business/energy-oil/google-nuclear-power-artificial-intelligence-87966624
Google Backs New Nuclear Plants to Power AI
Startup Kairos Power plans to build small reactors to help supply electricity to the tech company’s data centers, in a first-of-its-kind deal in the U.S.