AI chatbots have skyrocketed in popularity over the past few months, thanks to the growing success of OpenAI’s revolutionary ChatGPT, which surprisingly only hit the scene in December. But when Microsoft seized the opportunity to hitch its van to a whopping $10 billion OpenAI rising star, it decided to do so by introducing a GPT-4-based chatbot masquerading as Bing, its powerful yet capable search engine. engine, in an attempt to reverse Google’s dominance in search. Google quickly followed suit with its own homegrown Bard AI and unveiled plans to put AI responses ahead of traditional search results, a completely monumental change in one of the most important places on the web.

Both are advertised as experiments. And these “AI chatbots” are truly remarkable achievements: I’ve spent many nights with my kids happily creating the fantastic artwork of your dreams with Dall-E Bing Chat integration and causing sick raps about wizards who think lizards are the source of all magic and see them come to life in a matter of moments with these fantastic tools. I love them

But Microsoft and Google marketers were wrong. AI chatbots like ChatGPT, Bing Chat, and Google Bard should never be confused with search engines, much less their power. They’re more like those crypto-bros littering the comments on Elon Musk’s terrible new Twitter, yelling out truthful claims loudly and confidently, which in fact are often full of absolute bullshit.

These so-called “AI chatbots” do a fantastic job of synthesizing information and providing interesting, often accurate details about what you’re asking. But under the hood, they are actually large language models (LLMs) trained on billions or even trillions data points – text – from which they learn to anticipate which words must further based on your request. AI chatbots are not smart at all. They use word association patterns to produce results that sound are plausible for your query, then formulate them specifically without having any idea whether these words are actually linked together true. Heck, Google’s artificial intelligence can’t even reproduce facts about Google products correctly.

I have no idea who originally coined the term, but the memes are correct: these chatbots are essentially auto-correcting on steroids. No reliable sources of information, such as the search engines they infiltrate, despite the fact that this association implies trust.

They are crap generators. They are crypto bro.

Further reading: ChatGPT vs. Bing vs. Bard: Which AI is better?

AI chatbots say the worst things

The signs appeared immediately. In addition to all the talk about experimentation, Microsoft and Google made sure to point out that these LLMs sometimes give inaccurate results (“hallucinations,” in AI techno-parlance). “Bing is powered by artificial intelligence, so surprises and bugs are possible,” Microsoft said in a statement. “Be sure to check the facts and share feedback so we can learn and improve!” This became apparent when reporters uncovered unfortunate inaccuracies in the brilliant launch presentations of both Bard and Bing Chat.

These lies suck when you use Bing and you know Googleare the two largest search engines in the world. But combining search engines with large language models has even deeper implications, as a recent Washington Post a report that reveals how ChatGPT OpenAI “made up a sexual harassment scandal and named a real law professor as the defendant,” as the headline aptly summed it up.

This is exactly what it looks like. But things are much worse because of how this hallucinatory “scandal” was discovered.

Brad Chacos/IDG

You must go read the article. This is both great and scary. In essence, law professor John Turley was contacted by a legal colleague who asked ChatGPT to compile a list of legal scholars guilty of sexual harassment. Turley’s name was on the list, along with a quote. Washington Post article. But Turley wasn’t accused of sexual harassment, and that’s mail article does not exist. The big-tongue model hallucinated her, likely due to Turley being interviewed by the legal press for publications such as The New York Times. mail.

“It was pretty intimidating,” Turley said. Mail. “A statement of this kind is incredibly harmful.”

You’re damn right, it is. Statement like What could ruin someone’s career, especially after Microsoft’s Bing Chat AI quickly started spreading similar allegations with Turley’s name in the news. “Now Bing So alleging that Turley was accused of sexually harassing a student during a school trip in 2018,” mailX Will Oremus tweeted. “The source for this claim is Turley’s own USA Today article on the ChatGPT false claim, as well as several other aggregations of his article.”

I would be furious and vehemently sue every company involved in defamatory statements made under the corporate banners of OpenAI and Microsoft. Oddly enough, the mayor of Australia threatened exactly this at about the same time that mail report published. “Regional Australian Mayor [Brian Hood] said he could sue OpenAI if it didn’t correct ChatGPT’s false allegations that he had served time in prison for bribery, in what would be the first defamation lawsuit against the automated text service, Reuters reported.

OpenAI’s ChatGPT is bearing the brunt of these lawsuits, perhaps because it’s at the forefront of “AI chatbots” and has been the fastest technology to adopt. Always. (Spitting out slanderous, hallucinatory statements doesn’t help.) But Microsoft and Google are doing just as much harm by associating chatbots with search engines. They are too inaccurate for that, at least at this stage.

Turley and Hood’s examples may be extreme, but if you spend any time playing with these chatbots, you’re bound to stumble upon more insidious inaccuracies, nonetheless stated with complete certainty. Bing, for example, misidentified my daughter’s gender when I asked about her, and when I had him put together a personalized resume from my LinkedIn profile, he got a lot of the right but also hallucinatory skills from previous employers. This can be devastating to your job prospects if you don’t pay attention to it. Again, the bard reveal demonstration included evident a lie about the James Webb Space Telescope that astronomers immediately identified. Using these supposedly search-engine-friendly research tools can ruin your child’s school grades.

It wasn’t supposed to be like this

Bing Chat / Brad Chacos / IDG

The hallucinations that these AI tools sometimes spit out aren’t as painful in more creative endeavors. AI Art Generators skirt, as well as killer-looking Microsoft Office AI enhancements that can create full-fledged PowerPoint presentations from reference docs you cite, and more, seem poised to make drastic improvements to desktop drones like yours truly. But these tasks don’t have the strict accuracy expectations that come with search engines.

It shouldn’t have been like that. Microsoft and Google marketing really got it wrong by associating big language models with search engines in the eyes of the public, and I hope this doesn’t lead to permanent poisoning of the well of perception. These are fantastic tools.

I will end this article with a tweet from Steven Sinofsky who replied to a comment about severely misguided ChatGPT hallucinations causing a headache in an inaccurately cited researcher. Sinofsky is the investor who brought Microsoft Office and Windows 7 to fame in his time, so he knows what he’s talking about.

“Imagine a world where it’s called ‘Creative Writer’ and not ‘Search’ or ‘Ask anything about the world,'” he said. “Now it’s just a fiasco with branding. Maybe in 10 years of progress, many more layers of technology and so on will come to search.”

However, at the moment, AI chatbots are crypto bro. Have fun, enjoy the possibilities that these wonderful tools open up, but do not take their information at face value. This is true, not trustworthy.

Editor’s Note: This article was originally published on April 7, 2023, but was updated on May 12 after Google announced plans to place AI responses at the top of search results.