But you get 10X the nonsense in return.

They'll just have to run 10x the ads, which is fine since AdBlock still works.

 

They'll just have to run 10x the ads, which is fine since AdBlock still works.

You need to have ChatGPT ad placement directly in the results.
https://www.forbes.com/sites/lancee...uishing-ai-ethics-and-ai-law/?sh=2c25bcdd5fd5
Monetizing Generative AI ChatGPT Via Embedded Product Placement Ads, Anguishing AI Ethics And AI Law

 

You need to have ChatGPT ad placement directly in the results.
https://www.forbes.com/sites/lancee...uishing-ai-ethics-and-ai-law/?sh=2c25bcdd5fd5
Monetizing Generative AI ChatGPT Via Embedded Product Placement Ads, Anguishing AI Ethics And AI Law

There goes the neighborhood.

Back to Duck Duck Go, I guess.

 

There goes the neighborhood.

Back to Duck Duck Go, I guess.

 

https://finance.yahoo.com/news/companies-already-replacing-workers-chatgpt-140000856.html
Some companies are already replacing workers with ChatGPT, despite warnings it shouldn’t be relied on for ‘anything important’

As gung ho as business leaders appear to be about the potential of ChatGPT, it's not without its critiques, including concerns regarding cheating and plagiarism, racism and sexism bias, accuracy, and overall questions about how it's been trained to learn. The Atlantic's Ian Bogost warned it should be treated as a toy not a tool, and New York Times tech columnist Kevin Roose said that Microsoft's new A.I. version of its Bing search engine powered by ChatGPT's OpenAI left him feeling “deeply unsettled” and “even frightened” after a two-hour chat in which it sounded unhinged and somewhat dark.

OpenAI CEO Sam Altman has previously warned that ChatGPT shouldn't be relied on for "anything important," and in a recent series of tweets expressed concerns about the dangers posed by A.I. technology—and the iterations to follow—saying he worried how people of the future will view us.

 

Some companies are already replacing workers with ChatGPT, despite warnings it shouldn’t be relied on for ‘anything important’

To be fair, some worker don't do 'anything important.'

 

https://finance.yahoo.com/news/companies-already-replacing-workers-chatgpt-140000856.html
Some companies are already replacing workers with ChatGPT, despite warnings it shouldn’t be relied on for ‘anything important’

While extremely disappointing, it certainly isn't surprising. It's understandable that the "average" person doesn't have a clue about how piss-poor these AI tools are. I would certainly have a hard time spotting quality from pure bullcrap between two papers on something that I knew nothing about -- such as wine tasting -- as long as both were written in coherent sentences.

But you would sure hope and think that people making decisions about using it for work product in THEIR area of expertise would be able to spot the pure crap that it produces after even a superficial evaluation -- and certainly after the kind of in-depth evaluation that should be undertaken before making this kind of a decision.

But then, I long ago came to the conclusion that many people in the kinds of positions where these decisions get made are actually pretty incompetent in their supposed area of expertise. Instead, they tend to be the kind of people that value good sounding fluff over accurate substance and have a history of excelling at it in order to advance to those positions in the first place.

 

To be fair, some worker don't do 'anything important.'

No one is so completely useless that they cannot serve as a bad example. An adage I came up with to describe one of our fellow employees before he was fired for cause. As we say down here, he was as useless as teats on a boar hog.

 

​

 

https://futurism.com/the-byte/ftc-warns-keep-ai-claims-in-check

The US Federal Trade Commission (FTC) is well aware of AI's big boom, and it's not amused by some of the "breathless" hype surrounding it, particularly from the companies that peddle the technology.

On Monday, the agency issued a strongly worded statement directed at Silicon Valley heavyweights and the countless other businesses that tout AI to customers, reminding them that "false or unsubstantiated claims about a product's efficacy" is the FTC's "bread and butter."
...
One of the main concerns of the FTC is whether businesses are exaggerating what their AI products can do, such as deceptive performance claims that "lack scientific support" or only apply to "certain types of users or under certain conditions."

Another no-go: claiming an AI product does it better than a non-AI one.

"You need adequate proof for that kind of comparative claim, too, and if such proof is impossible to get, then don't make the claim," the agency stated.

And lastly, businesses need to be careful that their "AI" product actually uses AI.

 

https://futurism.com/the-byte/ftc-warns-keep-ai-claims-in-check

If only we could get them to send a similarly "strongly worded statement" to supplement peddlers and a host of other industries.

 

Its response was another lie with that canned text, obviously preprogrammed as a response to those types of questions. It's a bad joke IMO. It repeats the lies and misinformation feed to it with no care to provide truthful answers.

Its really funny, if you call it for some of its nonsense a few times or catch it giving wrong information even when it has given the correct information previously it will give you the "too many requests, wait 1hr: message.

 

https://www.theregister.com/2023/03/02/openai_api_chatgpt_whisper/

The ChatGPT model family released on Wednesday, gpt-3.5-turbo, is offered at a price of $0.002 per 1,000 tokens (~750 words), which OpenAI says is 10x cheaper than previous GPT-3.5 models. The Whisper large-V2 model is priced at $0.006 / minute. There's also an open source version of the code, though OpenAI admits it can be hard to run.

Max Woolf, a data scientist, in an online post, observes that that the API pricing is extraordinarily low.

“I have no idea how OpenAI can make money on this,” he said. “This has to be a loss-leader to lock out competitors before they even get off the ground.”

 

https://hai.stanford.edu/news/human-writer-or-ai-scholars-build-detection-tool

Human Writer or AI? Scholars Build a Detection Tool
DetectGPT can determine with up to 95% accuracy whether a large language model wrote that essay or social media post.

But as Mitchell pondered Khazatsky’s question, he had the initial intuition that because even powerful LLMs have subtle, arbitrary biases for using one phrasing of an idea over another, the LLM will tend to “like” any slight rephrasing of its own outputs less than the original. By contrast, even when an LLM “likes” a piece of human-generated text, meaning it gives it a high probability rating, the model’s evaluation of slightly modified versions of that text would be much more varied. “If we perturb a human-generated text, it’s roughly equally likely that the model will like it more or less than the original.”

 

https://hai.stanford.edu/news/human-writer-or-ai-scholars-build-detection-tool

Human Writer or AI? Scholars Build a Detection Tool
DetectGPT can determine with up to 95% accuracy whether a large language model wrote that essay or social media post.

Love it AI versus AI… Battle of the Century

 

AI versus AI...

 

https://hai.stanford.edu/news/human-writer-or-ai-scholars-build-detection-tool

Human Writer or AI? Scholars Build a Detection Tool
DetectGPT can determine with up to 95% accuracy whether a large language model wrote that essay or social media post.

So then all the ChatGPT people need to do is tack DetectGPT onto the end and export only the things that it "likes". Or incorporate it into the training sets so that it learns to generate things that DetectGPT will favor.

I think there is still the underlying problem -- neither system is making any serious attempt to validate the correctness of the gibberish involved, only making probabilistic evaluations of whether the each word should follow the one before it based on all of the samples of what words follow what words that it has previously seen.

 

So then all the ChatGPT people need to do is tack DetectGPT onto the end and export only the things that it "likes". Or incorporate it into the training sets so that it learns to generate things that DetectGPT will favor.

I think there is still the underlying problem -- neither system is making any serious attempt to validate the correctness of the gibberish involved, only making probabilistic evaluations of whether the each word should follow the one before it based on all of the samples of what words follow what words that it has previously seen.

That's the underlying problem but detecting the writing gait of ChatGPT as a 'tell' helps chose where to look closer like we do with humans.



It will be interesting to see if AI companies try to include specific measures to defeat ChatGPT “watermark” detection that could lead to undetectable malicious activity.
https://arxiv.org/pdf/2204.06974.pdf
Planting Undetectable Backdoors
in Machine Learning Models

Scott Aaronson wrote about how watermarking works:

“My main project so far has been a tool for statistically watermarking the outputs of a text model like GPT.
Basically, whenever GPT generates some long text, we want there to be an otherwise unnoticeable secret signal in its choices of words, which you can use to prove later that, yes, this came from GPT.”

 

​

 

https://www.wired.com/story/gpt-4-openai-will-make-chatgpt-smarter-but-wont-fix-its-flaws/
GPT-4 Will Make ChatGPT Smarter but Won't Fix Its Flaws

But that underlying mechanism also means that ChatGPT and systems like it will often make up facts. And despite OpenAI’s efforts to make the model resistant to abuse, it can be prompted into misbehaving, for example by suggesting it role-play doing something it refuses to do when asked directly. OpenAI says GPT-4 is 40 percent more likely to provide “factual responses” and says that GPT-4 is 82 percent less likely to respond to requests that should be disallowed. The company did not say how often the previous version, GPT-3, provides factually incorrect responses or responds to requests it should reject.

Still, Ilya Sutskever, cofounder and chief scientist at OpenAI, claims those as perhaps the most significant advances with the new model. “The thing that stands in the way of ChatGPT being really useful to many people for many tasks is reliability,” he says. “GPT-4 isn't there yet, but it is a lot closer.”

Conitzer at CMU says GPT-4 appears to include new guardrails that prevent it from generating undesirable responses but adds that its new capabilities may lead to new ways of exploiting it.