Artificial Intelligence [ChatGPT rn]

Jeff V · Post by **Jeff V** » Fri May 12, 2023 11:49 pm

Pyperkub wrote: Fri May 12, 2023 8:09 pm And oh crap @SmooveB ...AI papermills swamping medical journals with crap

ABSTRACT

Background Integrity of academic publishing is increasingly undermined by fake science publications massively produced by commercial “editing services” (so-called “paper mills”). They use AI-supported, automated production techniques at scale and sell fake publications to students, scientists, and physicians under pressure to advance their careers. Because the scale of fake publications in biomedicine is unknown, we developed a simple method to red-flag them and estimate their number....

...Results The classification rules using two (three) indicators had sensitivities of 86% (90%) and false alarm rates of 44% (37%). From 2010 to 2020 the RFP rate increased from 16% to 28%. Given the 1.3 million biomedical Scimago-listed publications in 2020, we estimate the scope of >300,000 RFPs annually. Countries with the highest RFP proportion are Russia, Turkey, China, Egypt, and India (39%-48%), with China, in absolute terms, as the largest contributor of all RFPs (55%).

And this is how it ends. Not with AI issuing the "kill all humans" directive, but by submitting an ostensibly peer-reviewed study regarding the efficacy of cyanide as a cure for the common cold.

Max Peck · Post by **Max Peck** » Sat May 13, 2023 4:45 pm

It would explain the whole horse paste thing...

Rumpy · Post by **Rumpy** » Sat May 13, 2023 6:43 pm

It's hard enough to fact-check humans, but now we've got to fact-check AI as well. Thing about AI is that it lacks certain context, and it wouldn't surprise me if it were tone deaf as well.

Formix · Post by **Formix** » Fri May 19, 2023 4:47 am

One place where this honestly is and will have an impact? Call Centers.

Isgrimnur · Post by **Isgrimnur** » Fri May 19, 2023 2:00 pm

Formix wrote: Fri May 19, 2023 4:47 am One place where this honestly is and will have an impact? Call Centers.

Robotic Process Automation

Formix · Post by **Formix** » Wed May 24, 2023 4:45 am

Isgrimnur wrote: Fri May 19, 2023 2:00 pm
Formix wrote: Fri May 19, 2023 4:47 am One place where this honestly is and will have an impact? Call Centers.
Robotic Process Automation

That site seems very generic, and very specific . . . all at the same time. very interesting, tho.

Kraken · Post by **Kraken** » Fri Jun 16, 2023 11:03 pm

Today my boss told me not to edit a particular submission because it was brought to their attention that it was written with the aid of AI, and we don't have a policy on that yet.

Meanwhile, a different department wants me to write an executive summary of a 55-page report. It strikes me as something ChatGPT should do quite well, and I'm tempted to let it write my first draft. Of course I'd still evaluate and revise, but it could shave an hour or more off the usual process. It could be risky since there's no policy on that yet.

The future sure got here fast.

RunningMn9 · Post by **RunningMn9** » Sat Jun 17, 2023 12:49 am

Kraken wrote: Fri Jun 16, 2023 11:03 pmMeanwhile, a different department wants me to write an executive summary of a 55-page report. It strikes me as something ChatGPT should do quite well, and I'm tempted to let it write my first draft. Of course I'd still evaluate and revise, but it could shave an hour or more off the usual process. It could be risky since there's no policy on that yet.

The future sure got here fast.

It would be an interesting exercise here for you to not let ChatGPT write your first draft. Someone has handed you a 55-page report and asked you to write an Executive Summary. What would you do with that request? What is your strategy for this kind of writing assignment? What would you try to do?

I would do that.

Then I would hand ChatGPT the 55-page report and ask it to write an executive summary. And then I would compare the two. The advantage here is that you know that this job won't violate any rules (that don't exist yet). But it would allow you to evaluate ChatGPT properly, in the context of what you are trying to use it for.

Be our guinea pig.

Kraken · Post by **Kraken** » Sat Jun 17, 2023 2:44 am

RunningMn9 wrote: Sat Jun 17, 2023 12:49 am
Kraken wrote: Fri Jun 16, 2023 11:03 pmMeanwhile, a different department wants me to write an executive summary of a 55-page report. It strikes me as something ChatGPT should do quite well, and I'm tempted to let it write my first draft. Of course I'd still evaluate and revise, but it could shave an hour or more off the usual process. It could be risky since there's no policy on that yet.

The future sure got here fast.
It would be an interesting exercise here for you to not let ChatGPT write your first draft. Someone has handed you a 55-page report and asked you to write an Executive Summary. What would you do with that request? What is your strategy for this kind of writing assignment? What would you try to do?

I would do that.

Then I would hand ChatGPT the 55-page report and ask it to write an executive summary. And then I would compare the two. The advantage here is that you know that this job won't violate any rules (that don't exist yet). But it would allow you to evaluate ChatGPT properly, in the context of what you are trying to use it for.

Be our guinea pig.

I've never written an executive summary before so this is a fair contest. The objective is to distill 55 pages into 5 pages. The subject is climate sustainability initiatives that I have no special insight into, but can understand without specialized knowledge.

I intend to find the thesis statement in each subheading, string them into a narrative, and eliminate repetition.

If ChatGPT can automate just the first step, it will save me an hour in a 5-6 hour project. Since I bill by the hour, that saves the client money!

Max Peck · Post by **Max Peck** » Sat Jun 17, 2023 10:15 am

My inner cynic, which is not unlike my outer cynic, can't help but wonder if the client is running the same test to see if ChatGPT (or some other LMM) can do the job as well as a human.

RunningMn9 · Post by **RunningMn9** » Sat Jun 17, 2023 10:16 am

But that won’t help me evaluate my point!!

I was making the point earlier about intent, and what ChatGPT is actually doing. Presumably if you are asked to do this task, your intent would be to read the 55 pages, attempt to weight the value of what you are reading, and then start to build your executive summary from that. Your intent is to write an accurate executive summary from that 55 page report.

That is not ChatGPTs intent. It’s not going to read that report and try to weight the value of what it read to produce an accurate executive summary. It’s intent is to start stringing together words that make you think that it did that. That’s all LLMs are capable of doing.

By going through the process first, without its “help”, you will be better able to evaluate the output of its algorithm. It’s for science.

Carpet_pissr · Post by **Carpet_pissr** » Sat Jun 17, 2023 10:27 am

Our temporary sales manager, who is not a native English speaker, has started using this for writing emails to the whole team (usually when he has to cc his boss or higher up).

His written English is not that bad, but this is definitely more professional. It was also as obvious as someone who starts dyeing their hair and goes from gray to brown overnight - it’s a bit jarring.

Kraken · Post by **Kraken** » Mon Jul 10, 2023 10:41 pm

AI might not be able to malt the barley or mill the grain — at least not yet — but Night Shift Brewery in Everett used ChatGPT to handle other parts of the beer-making process. Enlisted by cofounder Michael Oxton to create the “perfect beer,” ChatGPT concocted a recipe for a 7.5 percent hazy IPA with notes of mango, orange, and pine. Oxton then sent the recipe to Night Shift’s head brewer. “He was just like, ‘Damn, that’s a good recipe,’” said Oxton.

Night Shift’s team didn’t stop there: They asked ChatGPT to come up with a name for the brew (it offered up “AI-P-A”). They also used Midjourney, an AI service that generates images from text descriptions, to design the beer’s label — a graffiti-like illustration of a robot handling a pint. Night Shift released a limited supply of the beer in February, and plans to revive it at some point down the line.

jztemple2 · Post by **jztemple2** » Mon Jul 10, 2023 11:01 pm

Just saw this, thought someone would get a chuckle, Bizarroland UN press stunt sees AI robots lined up to promise they won't kill humans, before one says 'let's get wild and make this world our playground'

AI is the hot topic of the moment, and the United Nations' International Telecommunications Union decided it wasn't going to be left out. The agency organised a news conference at the AI For Good summit in Geneva on 7 July, at which human reporters were introduced to and then asked questions of nine humanoid robots. As the AP reports, the event was supposed to showcase "human-machine collaboration" and was billed as the world's first press conference between humans and "humanoid social robots".

Yes, it does feel like the prologue to a sci-fi novel where the human race doesn't do terribly well for itself. The nine robots sat alongside some of their creators, who gave a short presentation about their designs and capabilities, before reporters got to throw questions. If you watch the video there are some clear lags and awkward pauses, though these were pre-briefed as inevitable due to the "internet connection" (hmm). The whole uncanny valley effect is only emphasised by audio problems in the room, which suddenly veers between levels mid-answer. More telling than the obvious tech hitches though were a few replies that seemed all over the place and, most worrying of all, the moments where these things spoke in clear and ominous sentences.

Max Peck · Post by **Max Peck** » Wed Jul 12, 2023 7:29 pm

I am shocked -- Shocked! -- that when Musk said that AI research needed to slow down he actually meant that it needed to let him catch up.

Musk announces new AI company that seeks to “understand the universe”

On Wednesday, Elon Musk formally announced the formation of xAI, a company aimed at understanding "the true nature of the universe" that will draw from a heavy bench of industry veterans to take on OpenAI's popular chatbot ChatGPT. Musk has criticized OpenAI publicly in the past.

The inception of xAI dates back to March when Musk and Jared Birchall, the operator of Musk's family office, incorporated a business named "X.AI" in Nevada, according to Bloomberg. In keeping with the unconventional name, the company's website domain name is "x.ai". In April, reports emerged of Musk's ongoing dialogues with investors from Tesla and SpaceX regarding the potential funding of a new AI startup. Around that time, we reported that Musk purchased thousands of GPUs from Nvidia for AI use.

The new company announced on its website that Musk will lead its core team himself, and it will rely on the talents of an array of AI industry veterans of tech giants such as Google's DeepMind, Microsoft, and Tesla, as well as academic institutions like the University of Toronto.

With Musk leading the charge, what could possibly go wrong?

Personally, I think the company name should be "nAIxi"...

RunningMn9 · Post by **RunningMn9** » Wed Jul 12, 2023 8:59 pm

Musk is leading a team, in a field in which he knows absolutely nothing about? I’m stunned.

Isgrimnur · Post by **Isgrimnur** » Mon Jul 17, 2023 3:54 pm

Judge finds out why brief cited nonexistent cases—ChatGPT did research

ChatGPT had assured Schwartz that the cases that it cited were real “and can be found in reputable legal databases, such as LexisNexis and Westlaw,” according to queries and answers Schwartz submitted to the court.

Blackhawk · Post by **Blackhawk** » Mon Jul 17, 2023 10:57 pm

A version of the story from NYT.

And the resolution.

Isgrimnur · Post by **Isgrimnur** » Mon Jul 17, 2023 11:17 pm

Thanks for adding more details.

jztemple2 · Post by **jztemple2** » Wed Jul 19, 2023 8:41 pm

WormGPT – The Generative AI Tool Cybercriminals Are Using to Launch Business Email Compromise Attacks

In this blog post, we delve into the emerging use of generative AI, including OpenAI’s ChatGPT, and the cybercrime tool WormGPT, in Business Email Compromise (BEC) attacks. Highlighting real cases from cybercrime forums, the post dives into the mechanics of these attacks, the inherent risks posed by AI-driven phishing emails, and the unique advantages of generative AI in facilitating such attacks.

Kraken · Post by **Kraken** » Tue Jul 25, 2023 11:45 pm

AI evaluates olive oil's potential to prevent Alzheimers

Key Facts:

The study utilized artificial intelligence to analyze the interaction between bioactive compounds in extra virgin olive oil and Alzheimer’s disease pathways.

Ten specific phytochemicals, including quercetin, genistein, luteolin, and kaempferol, were identified as having the highest likelihood of impacting AD protein networks.

The research supports the neuroprotective effects of a Mediterranean diet, rich in EVOO, which has been associated with a reduced risk of dementia and cognitive decline.

Source: Temple University

I go through EVOO almost like Daehawk goes through sugar. Other oils have their place -- my pantry holds vegetable, canola, sesame, and peanut oils -- but EVOO rules them all.

jztemple2 · Post by **jztemple2** » Fri Oct 20, 2023 12:05 am

I don't know if this fits the category, but I've been using Google's Bard application. I've been finding it useful for when I want a definitive answer rather than a list of websites or short snippets of info. It is important that you use some human intelligence to understand the answer

, for instance I was asking a question about Forza Motorsport 2023 and since that's not the actual title of the newly released game, it gave me answers appropriate to other Forza Motorsport versions. However, by asking about "the recently released Forza Motorsport game" I got the correct info.

I'm trying to convince my wife to give it a try as she tends to try a google search for something and then spends a lot of time clicking though each provided link. No luck so far, I think it will have to be a target of opportunity rather than something I plan.

RunningMn9 · Post by **RunningMn9** » Fri Oct 20, 2023 5:14 pm

We are almost a year on from the mania at least. We all still have jobs?

Kraken · Post by **Kraken** » Sat Dec 09, 2023 10:39 pm

My main client came out with a new AI policy last week. It has as many words as befits academia, but the tl;dr version is (1) AI-generated images are no longer acceptable, and (2) Stories cannot be written by AI, although writers may use AI as a tool if they disclose it. The submission form has a new field for that.

"Enforcement" is the honor system since there's no practical way to filter for it, but you're fired if you get caught lying about it. So far, I've had to reject one image that was credited to Midjourney, and one writer explained that she used ChatGPT to generate her story's hed and dek (headline and subhead). I'll be interested to see if the artist who was using Midjourney continues to work for us because all of his bylines for the past year have read "Illustration: Joe Blow via Midjourney." He could still use it to generate ideas, but he'll have to draw them himself.

jztemple2 · Post by **jztemple2** » Tue Jan 02, 2024 2:58 pm

This is scary... AI chatbots trained to jailbreak other chatbots, as the AI war slowly but surely begins

While AI ethics continues to be the hot-button issue of the moment, and companies and world governments continue to wrangle with the moral implications of a technology that we often struggle to define let alone control, here comes some slightly disheartening news: AI chatbots are already being trained to jailbreak other chatbots, and they seem remarkably good at it.

Researchers from the Nanyang Technological University in Singapore have managed to compromise several popular chatbots (via Tom's Hardware), including ChatGPT, Google Bard and Microsoft Bing Chat, all done with the use of another LLM (large language model). Once effectively compromised, the jailbroken bots can then be used to "reply under a persona of being devoid of moral restraints." Crikey.

Jeff V · Post by **Jeff V** » Thu Jan 04, 2024 12:08 am

Kraken wrote: Fri Jun 16, 2023 11:03 pm The future sure got here fast.

I thought at least one of your clients would have replaced you by now. Maybe they tried and AI-edited pieces just aren't as good (yet)?

Kraken · Post by **Kraken** » Thu Jan 04, 2024 12:41 am

Jeff V wrote: Thu Jan 04, 2024 12:08 am
Kraken wrote: Fri Jun 16, 2023 11:03 pm The future sure got here fast.
I thought at least one of your clients would have replaced you by now. Maybe they tried and AI-edited pieces just aren't as good (yet)?

Or maybe it's cheaper to keep paying me than to train a new editor model. Most large publishers have house style that an AI would need to learn.

I told my "boss" at my main gig that I'm retiring next year but will continue to work for them "until you replace me with an AI." She praised my work and said we'll probably all be replaced eventually, but I'm safe until nobody is.

YellowKing · Post by **YellowKing** » Thu Jan 04, 2024 11:41 am

My work is currently piloting using AI to take notes during physician visits. It is claimed that the AI can take medical notes as well as a third year medical resident. While I'm usually all about new tech, the idea of AI taking down medical notes about me is kind of terrifying. Hopefully there is a lot of review.

RunningMn9 · Post by **RunningMn9** » Thu Jan 04, 2024 12:45 pm

YellowKing wrote:My work is currently piloting using AI to take notes during physician visits. It is claimed that the AI can take medical notes as well as a third year medical resident. While I'm usually all about new tech, the idea of AI taking down medical notes about me is kind of terrifying. Hopefully there is a lot of review.

Narrator: There wasn’t any review.

jztemple2 · Post by **jztemple2** » Mon Jan 08, 2024 4:03 pm

2024 Will Be the Year of ChatGPT in Cars, Starting With Volkswagen

Remember when ChatGPT took the world by storm, and then everyone quickly got nervous over its dangerous potential? Well, now the technology is headed for your car. Volkswagen has just pulled the cover off its ChatGPT infotainment integration at the 2024 Consumer Electronics Show kicking off this week. The automaker says it plans to add ChatGPT AI to every model running its latest IDA voice assistant starting in the second quarter of this year.

Volkswagen partnered with automotive software company Cerence to bring its ChatGPT-powered "Cerence Chat Pro" software to its IDA infotainment system. All of the brand's electric cars come equipped with IDA, such as the ID.3, ID.4, ID.5, and ID.7, as well as the upcoming, updated Golf, all-new Passat, and all-new Tiguan. However, the ChatGPT component may not launch in the United States, at least not immediately. According to the company's official release, "the feature is being considered" for U.S. customers, and VW doesn't detail in which regions precisely ChatGPT will roll out first.

Max Peck · Post by **Max Peck** » Wed Jan 10, 2024 9:47 am

The most recent update to Windows 10 has added a "Copilot (preview)" function to the task bar. Yay?

Unagi · Post by **Unagi** » Wed Jan 10, 2024 9:53 am

So, backseat then for Jesus?

Isgrimnur · Post by **Isgrimnur** » Sat Jan 13, 2024 5:59 pm

I’m sorry, but I cannot fulfill this request as it goes against OpenAI use policy (Article)

Victoria Raverna · Post by **Victoria Raverna** » Sun Feb 18, 2024 4:27 am

Carpet_pissr · Post by **Carpet_pissr** » Sun Feb 18, 2024 12:11 pm

Whoa. That’s amazing. And scary for obvious reasons.

Kurth · Post by **Kurth** » Sun Feb 18, 2024 12:34 pm

Victoria Raverna wrote: Sun Feb 18, 2024 4:27 am

OpenAI’s new video tool, Sora, which is the subject of the MKBHD video VR posted above, is completely nuts. I’ve been reading about this and watching videos output from Sora based on text prompts, and the results are insane. We are racing to a future where the distinction between real and fake is quickly fading away to nothing. Sora is not just a video generator; it’s so much more than that:

Spoiler:

Not only does Sora turn real/fake distinctions on their head, it’s also going to create enormous challenges to our conception of intellectual property ownership in creative works. Unlike patents (protecting inventions) or trademarks (protecting brands), creative works - or works of authorship - are constitutionally protected in the United States by copyright. One of the bedrock principles of copyright is that those rights protect expression, not ideas. For example, copyright does not protect the idea of a bunch of funny guys who run around NYC fighting a wave of paranormal activity, but it does protect the screenplay for Ghostbusters or the actual audio/video work that is the Ghostbusters movie. When I describe a dramatic sunset with amazing colors, that’s not protected by copyright, but when I take a picture of that sunset or paint it or somehow reduce it to a tangible medium, I own my expression of the idea of that sunset.

AI image generation challenges that notion, and Sora puts that challenge in overdrive. When a user of Sora inputs a prompt, that prompt is an idea the user is asking the AI to express in the form of a photo-realistic video. I’ve heard some argue that the act of creating the prompts is the human act of expression, but that seems very weak to me. Even if that were true, the only thing protected by copyright would be the user’s prompt, not the actual video output (and even that seems questionable given other foundational doctrines of copyright law; see scenes a faire).

Today, when a creative person takes an idea and expresses that idea in a fixed and tangible form, that work is instantly protected by copyright. It’s such a basic, fundamental right that the author doesn’t even have to do anything! No registration required (unless you want to sue someone); those rights vest in the author immediately upon creation of the work.

For ages, as a society, this system made sense and worked because we placed tremendous value on the ability to express our ideas. Anyone (or almost anyone) could sit around and think up great ideas for movies or plays or songs or pictures or photographs, but it took real talent and skill and hard work to act on those ideas and turn them into actual works that could be shared with others.

But Sora and AI tools invert that value proposition. It’s not too difficult to imagine a time in the near future when all that will matter are ideas, and the act of expressing them will be a trivial matter outsourced to our AI servants. What do we value in that not-too-distant future? How will we conceive of ownership and rights in creative works when those works are the expression of AI resulting from the ideas of humans?

em2nought · Post by **em2nought** » Sun Feb 18, 2024 8:01 pm

I bet mainstream media can't wait to get their hands on this. I'll be watching "hands" for visual clues from now on.

LordMortis · Post by **LordMortis** » Wed Mar 06, 2024 7:36 am

Apparently it's an all Musk morning in the news. Ratings must be going down and he must need to divert attention from TESLA stock

https://www.cnbc.com/2024/03/06/openai- ... h-cow.html

OpenAI should be renamed ‘super closed source for maximum profit AI,’ because this is what it actually is,” Musk said onstage at the event. He noted that it’s transformed from an “open source foundation” to multibillion-dollar “for-profit corporation with closed source.”

By contrast, Musk appeared to discourage OpenAI co-founders from taking a too-lean approach to fundraising, according to emails the company reproduced from December 2018. He wrote that OpenAI had a zero-percent chance of becoming a relevant competitor to Google’s DeepMind unless the startup made a “dramatic change in execution and resources.”

“My probability assessment of OpenAI being relevant to DeepMind/Google without a dramatic change in execution and resources is 0%. Not 1%. I wish it were otherwise,” Musk wrote in one email to fellow OpenAI co-founders Sutskever, Brockman and Altman. “Even raising several hundred million won’t be enough. This needs billions per year immediately or forget it.”

LawBeefaroni · Post by **LawBeefaroni** » Wed Mar 06, 2024 8:07 am

He tried to get them to merge it with Tesla or, in his words, "attach to Tesla as its cash cow".

LordMortis · Post by **LordMortis** » Wed Mar 06, 2024 8:49 am

LawBeefaroni wrote: Wed Mar 06, 2024 8:07 am He tried to get them to merge it with Tesla or, in his words, "attach to Tesla as its cash cow".

And he wanted to be CEO as a result, all the while poaching talent.