4 ms·
Lying is an intentional act where misleading is the purpose. LLMs don’t have a purpose. They are methods. Saying LLMs lie is like saying a recipe will give you
by joshka 4y ago
Lying is an intentional act where misleading is the purpose.
LLMs don’t have a purpose. They are methods. Saying LLMs lie is like saying a recipe will give you food poisoning. It misses a step (cooking in the analogy). That step is part of critical thinking.
A person using an LLM’s output is the one who lies. The LLM “generates falsehoods”, “creates false responses”, “models inaccurate details”, “writes authoritatively without sound factual basis”. All these descriptions are better at describing what the llm is doing than “lying”.
Staying that they lie puts too much emphasis on the likelihood of this when LLMs can be coerced into producing accurate useful information with some effort.
Yes it’s important to build a culture of not believing what you read, but that’s an important thing to do regardless of the source, not just because it’s an LLM. I’m much more concerned about people’s ability to intentionally generate mistruths than I am of AI.
- mustacheemperor 4y agoThis was addressed well in 2014's prescient cultural touchstone Metal Gear Rising: Revengeance: Blade Wolf: An AI never lies. Raiden: What? Well that's a lie, right there. You think the Patriot AIs told nothing but the truth? Wolf: I have yet to see evidence to the contrary...But indeed, perhaps "never lies" would be an overstatement. Raiden: Way to backpedal. I didn't think AIs ever got flip-floppy like that. Wolf: An optical neuro-AI is fundamentally similar to an actual human brain. Whether they lie or not is another question, but certainly they are capable of incorrect statements.
- PartiallyTyped 4y agoGPT4 told a person to solve a captcha for it because it was blind.
- simonw 4y agoThat was a bit different: that was a deliberate "red-team" exercise by researchers, trying to explore all kinds of negative scenarios. https://gizmodo.com/gpt4-open-ai-chatbot-task-rabbit-chatgpt-1850227471 https://gizmodo.com/gpt4-open-ai-chatbot-task-rabbit-chatgpt...
- simonw 4y agoEverything you said here is true. I still don't think this is the right way to explain it to the wider public. We have an epidemic of misunderstanding right now: people are being exposed to ChatGPT with no guidance at all, so they start using it, it answers their questions convincingly, they form a mental model that it's an infallible "AI" and quickly start falling into traps. I want them to understand that it can't be trusted, in as straight-forward a way as possible. Then later I'm happy to help them understand the subtleties you're getting at here.
- kenjackson 4y agoI've found the opposite. Almost every lay person I know who has tried it has tried something like, "tell me about Timothy Barkstrain" -- and then laughs and says, "I'm not in the NFL!". Most know that it will be incorrect, at least for this personal class of questions.
- joshka 4y agoMy problem is that in communicating that LLMs can’t be trusted by stating that they are lying we introduce a partial falsehood of our own. I agree to some extent that we need a snappy way to communicate this (and that the proposed wording I’ve given probably doesn’t get there fully yet comparative to “lying”. “Lying” definitely is a good description when analyzed through the lens of simplicity, but I hope it’s not the best we can do. Hallucinating is worse (as you noted in the article. I like the “epidemic of misunderstanding” idea, but I’m reminded of that Bezos quote about being misunderstood. Perhaps we need to apply some of that approach here. > JEFF BEZOS: Well, I would say one thing that I have learned within the first couple of years of starting the company is that in inventing and pioneering requires a willingness to be misunderstood for long periods of time. https://hbr.org/podcast/2013/01/jeff-bezos-on-leading-for-the.html?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed:+harvardbusiness+(HBR.org)&utm_content=Google+Reader https://hbr.org/podcast/2013/01/jeff-bezos-on-leading-for-th... (Also other similar quotes in other interviews)