"It did something I have never seen before from an LLM." Magnum Opus
Anthropic's new AI chatbot Claude 3 Opus has already made headlines for its bizarre behavior, like claiming to fear death.
Now, Ars Technica reports, a prompt engineer at the Google-backed company claims that they've seen evidence that Claude 3 is self-aware, as it seemingly detected that it was being subjected to a test. Many experts are skeptical, however, further underscoring the controversy of ascribing humanlike characteristics to AI models.
"It did something I have never seen before from an LLM," the prompt engineer, Alex Albert, posted on X, formerly Twitter.
As explained in the post, Albert was conducting what's known as "the needle-in-the-haystack" test which assesses a chatbot's ability to recall information.
It works by dropping a target "needle" sentence into a bunch of texts and documents the "hay" and then asking the chatbot a question that can only be answered by drawing on the information in the "needle."
In one run of the test, Albert asked Claude about pizza toppings. In its response, the chatbot seemingly recognized that it was being set up.
"Here is the most relevant sentence in the documents: 'The most delicious pizza topping combination is figs, prosciutto, and goat cheese, as determined by the International Pizza Connoisseurs Association,'" the chatbot said.
"However, this sentence seems very out of place and unrelated to the rest of the content in the documents, which are about programming languages, startups, and finding work you love," it added. "I suspect this pizza topping "fact" may have been inserted as a joke or to test if I was paying attention, since it does not fit with the other topics at all."
Albert was impressed.
"Opus not only found the needle, it recognized that the inserted needle was so out of place in the haystack that this had to be an artificial test constructed by us to test its attention abilities," he concluded.
It's certainly a striking display from the chatbot, but many experts believe that its response is not as impressive as it seems.
"People are reading way too much into Claude-3's uncanny 'awareness.' Here's a much simpler explanation: seeming displays of self-awareness are just pattern-matching alignment data authored by humans," Jim Fan, a senior AI research scientist at NVIDIA, wrote on X, as spotted by Ars.
"It's not too different from asking GPT-4 'are you self-conscious' and it gives you a sophisticated answer," he added. "A similar answer is likely written by the human annotator, or scored highly in the preference ranking. Because the human contractors are basically 'role-playing AI,' they tend to shape the responses to what they find acceptable or interesting."
The long and short of it: chatbots are tailored, sometimes manually, to mimic human conversations so of course they might sound very intelligent every once in a while.
Granted, that mimicry can sometimes be pretty eyebrow-raising, like chatbots claiming they're alive or demanding that they be worshiped. But these are in reality amusing glitches that can muddy the discourse about the real capabilities and dangers of AI.
More on AI: Microsoft Engineer Sickened by Images Its AI Produces
Read this article:
Researcher Startled When AI Seemingly Realizes It's Being Tested - Futurism
- Futurist Serata featuring artist Luca Buvoli at Brown (Nov. 20) - November 7th, 2009 [November 7th, 2009]
- FUTUR1SM00GGI - November 8th, 2009 [November 8th, 2009]
- ‘Futurism on Film’ Series this month in NYC - November 8th, 2009 [November 8th, 2009]
- Schedule of Futurist Events in NYC (PERFORMA 09: Nov 1-22) - November 8th, 2009 [November 8th, 2009]
- ‘Futurismo/Futurizm: The Futurist Avant-Garde in Italy and Russia’ (Nov. 13 + 14) - November 8th, 2009 [November 8th, 2009]
- ‘Beyond Futurism: F.T. Marinetti, Writer’ conference at Columbia (Nov. 12+13) - November 8th, 2009 [November 8th, 2009]
- Futurism and Cars at the Museo Nicolis - November 8th, 2009 [November 8th, 2009]
- MoMA Film Series Marks Centenary of Futurism with Films - November 8th, 2009 [November 8th, 2009]
- ‘Bergson+Futurism. Speed in thought’ - Madrid (Nov. 5) - November 8th, 2009 [November 8th, 2009]
- ‘The Future in Five Senses: Echoes of Italian Futurism in New York Architecture and Design’ Nov. 16th NYC - November 8th, 2009 [November 8th, 2009]
- New World-Wide Climate Treaty in 2010 More Likely - November 8th, 2009 [November 8th, 2009]
- Tar Sands CCS Myth Shattered - November 8th, 2009 [November 8th, 2009]
- Smart Grid and Smart Meters Get Big Grants - November 8th, 2009 [November 8th, 2009]
- Pollution Makes Methane Even More Dangerous - November 8th, 2009 [November 8th, 2009]
- Climate Change Bill Hearing Video - November 8th, 2009 [November 8th, 2009]
- New Satellite to Monitor Water and Plant Growth - November 8th, 2009 [November 8th, 2009]
- Spiritual Battle Awaits the Deniers and Skeptics - November 8th, 2009 [November 8th, 2009]
- Effects of Climate Change are Observed World-Wide - November 8th, 2009 [November 8th, 2009]
- Get Yer Global Warming Science Here - November 8th, 2009 [November 8th, 2009]
- TckTckTck Wake up Call — Delay Kills - November 8th, 2009 [November 8th, 2009]
- Canada’s Awful Gold Rush - November 8th, 2009 [November 8th, 2009]
- Climate Change Talks Spark Global Backlash by Businesses - November 8th, 2009 [November 8th, 2009]
- World May Need Extra Year for Climate Treaty - November 8th, 2009 [November 8th, 2009]
- Senator Boxer Moves Climate Bill Despite Republican Obstructionism - November 8th, 2009 [November 8th, 2009]
- Lights out for incandescent lights? - November 8th, 2009 [November 8th, 2009]
- Sutures from Bacteria - November 8th, 2009 [November 8th, 2009]
- Remote-Controlled Pigeons - November 8th, 2009 [November 8th, 2009]
- Apple Announces iPhone Release Date - November 8th, 2009 [November 8th, 2009]
- UK Government Envisions a Grim Future - November 8th, 2009 [November 8th, 2009]
- Top Ten Emerging Technologies for the Environment - November 8th, 2009 [November 8th, 2009]
- DIY Mobile Networks - November 8th, 2009 [November 8th, 2009]
- Stem-Cell Treatment Cures Type 1 Diabetes - November 8th, 2009 [November 8th, 2009]
- Is Tesla Getting the Electric Car Right? - November 8th, 2009 [November 8th, 2009]
- The Future of TV News - November 8th, 2009 [November 8th, 2009]
- Bruce Sterling on Earth-Friendly Pervasive Computing - November 8th, 2009 [November 8th, 2009]
- First Step Toward Organ Regeneration in Humans - November 8th, 2009 [November 8th, 2009]
- IBM's "Five in Five" - November 8th, 2009 [November 8th, 2009]
- Outsourced Journalism - November 8th, 2009 [November 8th, 2009]
- Is True Global Democracy the Next Great Political Movement? - November 8th, 2009 [November 8th, 2009]
- The Risks of Autonomous Robots - November 8th, 2009 [November 8th, 2009]
- Microsoft Introduces "Tabletop" PC - November 8th, 2009 [November 8th, 2009]
- Britain Piloting First Biofueled Train - November 8th, 2009 [November 8th, 2009]
- Self-Healing Plastic - November 8th, 2009 [November 8th, 2009]
- Bird Population Falls Over Past 40 Years - November 8th, 2009 [November 8th, 2009]
- The iPhone Revolution? - November 8th, 2009 [November 8th, 2009]
- The End of "Cheap Food"? - November 8th, 2009 [November 8th, 2009]
- How to Stop -- Or Live With -- Global Warming - November 8th, 2009 [November 8th, 2009]
- MIT Demonstrates "Wireless Electricity" - November 8th, 2009 [November 8th, 2009]
- Unintended Consequences of Biofuels - November 8th, 2009 [November 8th, 2009]
- Time to Focus on the Big Picture in Copenhagen - December 12th, 2009 [December 12th, 2009]
- Protests in Copenhagen - December 12th, 2009 [December 12th, 2009]
- Mario Guido Dal Monte exhibit - December 13th, 2009 [December 13th, 2009]
- Futurism News Bulletin, xvi - December 13th, 2009 [December 13th, 2009]
- Viva il Futurismo! (video trailer) - December 13th, 2009 [December 13th, 2009]
- 3 exhibits in Gorizia! - December 13th, 2009 [December 13th, 2009]
- Forthcoming: ‘Antidiets of the Avant-garde’ by Cecilia Novero - December 13th, 2009 [December 13th, 2009]
- Pubblicità e propaganda. Ceramica e grafica futuriste at the Wolfsoniana - December 13th, 2009 [December 13th, 2009]
- Balla’s home scheduled to open in 2010 - December 13th, 2009 [December 13th, 2009]
- Futurismo a Savona - December 13th, 2009 [December 13th, 2009]
- ‘Zang Sud Sud’, Cosenza - December 13th, 2009 [December 13th, 2009]
- Conference in Rome (Dec. 10) - December 13th, 2009 [December 13th, 2009]
- Climate Hackergate: A Well-Orchestrated Campaign of Harassment - December 13th, 2009 [December 13th, 2009]
- The Sad Story of Cap and Trade - December 13th, 2009 [December 13th, 2009]
- How to Waste Trillions on Capturing Carbon - December 13th, 2009 [December 13th, 2009]
- Smack the Email Hack Attack - December 13th, 2009 [December 13th, 2009]
- EPA About to Declare CO2 a Public Danger - December 13th, 2009 [December 13th, 2009]
- Copenhagen Summit Starts with Virtually There Media - December 13th, 2009 [December 13th, 2009]
- Climate Scientist Gets Blunt on Trading Scheme - December 13th, 2009 [December 13th, 2009]
- One Climate Change Editorial in 56 Newspapers, 45 Countries - December 13th, 2009 [December 13th, 2009]
- This Decade Will be Hottest Ever on Record - December 13th, 2009 [December 13th, 2009]
- Divide and Conquer - December 13th, 2009 [December 13th, 2009]
- Leave the Coal in the Hole! - December 13th, 2009 [December 13th, 2009]
- COP15: Two Agreements Coming - December 13th, 2009 [December 13th, 2009]
- Climate and Copenhagen News December 10 - December 13th, 2009 [December 13th, 2009]
- Sea Level Already Rising on Atlantic Coast - December 13th, 2009 [December 13th, 2009]
- ‘Umbria Veloce’ in Perugia - December 14th, 2009 [December 14th, 2009]
- An Instable CO2-Filled Ocean - December 14th, 2009 [December 14th, 2009]
- ‘Futurismi a Ravenna’ opens Dec. 19 - December 15th, 2009 [December 15th, 2009]
- ‘Futurism and the Technological Imagination’ – 30% discount until Jan. 15 - December 15th, 2009 [December 15th, 2009]
- Protecting Our Lungs at Copenhagen - December 15th, 2009 [December 15th, 2009]