BuzzFeed Stock Soars on News of Plans for AI to Create Content
“Instead of journalists learning to code, the code is learning journalism.”
Legal Insurrection readers may recall that I reported on ChatGPT (Chat Generative Pre-trained Transformer), a chatbot launched by OpenAI this past November. The New York Times hailed it as “the best artificial intelligence chatbot ever released to the general public.”
I also noted that the chatbot passed a Wharton Business School test as well as the US Medical Licensing exam . . . it is that good.
Now comes news that another program from OpenAI will be used to create content for the digital media and entertainment site BuzzFeed.
BuzzFeed Inc. said it would rely on ChatGPT creator OpenAI to enhance its quizzes and personalize some content for its audiences, becoming the latest digital publisher to embrace artificial intelligence.
In a memo to staff sent Thursday morning, which was reviewed by The Wall Street Journal, Chief Executive Jonah Peretti said he intends for AI to play a larger role in the company’s editorial and business operations this year.
In one instance, the company said new AI-powered quizzes would produce individual results.
…Mr. Peretti expects AI to assist the creative process and enhance the company’s content, while humans play the role of providing ideas, “cultural currency,” and “inspired prompts,” he wrote in his memo. In 15 years, he wrote, he expects AI and data to help “create, personalize, and animate the content itself,” rather than just curate existing content.
On that news, BuzzFeed stocks soared.
Shares of BuzzFeed Inc (BZFD.O) jumped on Thursday on reports of a deal with Meta Platforms Inc (META.O) and plans to use artificial intelligence to personalize and enhance the digital media firm’s online quizzes and content.
The stock was 19% higher in extended trading, after more than doubling in value earlier in the day as a Wall Street Journal report said it would use ChatGPT creator OpenAI for its content.
Buzzfeed, in an emailed response to Reuters, said “we are not using ChatGPT – we are using OpenAI’s publicly available API (application programming interface).”
A demonstration showed the power of a fully AI-generated BuzzFeed quiz.
In a private demonstration, BuzzFeed’s A.I.-powered quiz module wrote a paragraph for me about my own personalized secret society. I entered details like my name, a friend’s name and my most recent meal. (I picked pepperoni pizza.)
The A.I. spit back a detailed description of the society, including an initiation ritual that involved eating copious amounts of pepperoni pizza.
It was like a custom Mad Libs, if your book of Mad Libs were a bit more sentient. In this case, the A.I. — using a prompt created by a person — wrote a little story about a secret society with the words and proper names I’d supplied.
There is definitely an evolution in AI that is happening . . . just not the one the journalists were expecting, I am sure.
Instead of journalists learning to code, the code is learning journalism.
— Mouse Traditionalist Magnífico (@MagnificoIX) January 26, 2023
Donations tax deductible
to the full extent allowed by law.
Buzzfeed should name itself “B******tFeed”. How a company like this makes money is beyond comprehension. — Unless the likes of Soros, the CIA and/or Xi are undewriting it.
Hector Levesque and his Winograd Schema Challenge. AI score 60% versus winograd. Ordinary non-expert humans score 95% versus winograd.
The latest chatbot is pretty awesome. There’s also a lot of misleading industry hype. Investment hype. Bordering on tech-mobster pump-and-dump investment hype.
A recent book on computer intelligence ‘The Myth of Artificial Intelligence’ is an excellent and non-technical read and explains differences between Big Data and its narrow applications (chat, chess, go) and actual general artificial intelligence.
The acronym would more accurately stand for “Algorithm Implementation”. Human programmers have already mapped out the flowchart in advance, and the program just follows the directions. Computers are secretly dumb. They’re just very good at following directions, hundreds of thousands of them a second at times, and this creates an illusion of intelligence.
Creating a complex algorithm can be a satisfying intellectual challenge. But running that algorithm is so braindead easy, a computer can do it.
Sorry, but this response exhibits a complete misunderstanding of AI. It simply doesn’t work this way, and it was known more than 50 years ago, when I trained in it, that this approach would never produce the desired results.
Plus, “the computer” is not “the AI” — the programming network is. You may as well propose that souls can’t be complex at all, because they run on simple meat.
No, that’s not how Chat-GPT works. It is based on a neural network processor, and these types of algorithms do not require explicit instructions for every conceivable scenario. There is no ‘flow chart’, you ‘train’ the model by giving it textual information to consume.
But it is following an algorithm nonetheless. When they leave out the explicit instructions they get the expected results.
Example: the AI would conclude that USA is a white supremacist nation because it synthesizes a lot of text saying so. If you don’t want a racist computer you have to tell it something else.
“Neural nets” in AI, converge on generating the inputs to responses as provided in the “training data”, of inputs each paired with the “right” response, without telling you how they do it internally. (Mumble, mumble, hand-wave — you can’t get a set of descriptive language, propositions, or even “algorithm” out of a trained “Neural Net.” Different from one criteria of “Knowledge Discover in Databases — aka KDD.”, where the point is finding declarative knowledge that describes what you’ve seen. The first is folklore, the second is science.
A Neural Net trained up on mass, general population stimulus/response out in the Chat-o-Sphere would do well at generating expected hot takes, comfortable to most of the twits. So, a Colbert monologue generator. One understands why automating this would make BuzzFeed more valuable in it’s niche of … exactly that kind of regurgitation.
For every advance in artificial intelligence there is an equal advance in artificial stupidity.
Please! Only natural, free range organic cage-free stupidity is proper.
You forgot “fair-trade” and “cruelty-free.”
Content will be accompanied by proof-of-work, like bitcoin. It’s the labor theory of value.
It will improve the quality and accuracy of the reporting by leaps and bounds.
Yeah. My first reaction is that the most obvious improvement will be way more plausible lies. Which is timely indeed, because over the past four or so years, the Swamp has been practically phoning theirs in.
“Russian misinformation laptop.”
“Safe and effective.”
“I never talked with my son about his business.”
“Strongest economy ever.”
“You mean, with a cloth?”
But reporters, being imperfect, sometimes get it right. The machines will be programmed to be 100% wrong.
I’ve seen transcripts from this chatbot where its “certain answer” (which happens to align with the preferred narrative) turns 180* after a single prompt from the human, referencing a story, article, or other proof that it’s false. “You’re right, I had not been updated with that information…”
Interesting, not surprising use of “certain” by the bot and it’s wranglers.
How absurd for their stock price to be rising on this ‘news.’ There is absolutely no barrier to entry here… any [email protected] can do what they’re claiming they’re going to do.
This is boo.com all over again.
Maybe there is hope for USA finances after all. AI ought to be able to do better than at least half of Federal employees, including the current President and Vice-President.
The perfection of fake news that will be used to further manipulate people and create more dupes. How nice!