AI makes a mistake, surely not?

[LindorDoubleChoc]

I just googled a public figure because I wanted to know how old they are.

The first result on the page, the AI overview, said celeb is 70 years of age, having celebrated their 70th birthday on 16th April 2024.

Excuse me? Does AI often get it wrong, and how could that happen - anyone know?

Ai is wrong ALL the time! I don’t trust anything chat GpT says - I asked it who was in power in the UK last week and it said the conservatives?!! I don’t trust anything it says currently - I think it was better last year, now is in a weird phase where it’s grown exponentially but can’t keep up with sifting facts from fiction

Yes, it often gets things wrong. If it can’t find all the information to give a complete answer it will essentially make stuff up to fill in the gaps.

Are you being tongue in cheek, doesn't everyone know that AI isn't an oracle of truth?

Ask chatGPT when it's information is correct until.

My built-in knowledge goes up to June 2024. For anything after that, I can look things up on the web if you’d like me to. Do you want me to always check live sources when you ask about current events or things that change quickly?

It can also hallucinate and talk rubbish. I gave mine a reality filter directive. I will share it.

I find it quite disturbing that you think AI is 100% accurate 😳

AI is great for things like creating the first draft of a document if you give it the key points or creating a household budget.

I wouldn't ask it a factual question because you can't check what sources it has used. Info Out is only as good as Info In.

The worst that I regularly come across are the stupid AI captions that appear under FB posts.

Yep of course AI is incorrect. Sometimes it's right, but sometimes it's wrong. And this is the exact reason I will not use it, because you then have to fact check it therefore doing the work yourself.... you may as well just do the work yourself first off!

Have you been living under a rock or something? Surely everyone knows that AI regularly gets stuff wrong and/or makes things up.

The only surprising thing is that anyone would expect it to be accurate.

Do you? Why is that? I don't have any conscious experience of AI (apart from the hundreds of made up threads on Mumsnet), I've never knowingly sought an answer through AI or used it for anything.

Also ... I thought it was interesting because the answer to the question was a solid fact. AI wasn't being asked to be nuanced or creative. Infact, it gave two contradictory answers to the one question!

I love watching ai gone wrong videos on youtube. Theres loads of them. I once asked how many Saturdays until Christmas.

I googled to fact check something I thought was wrong before I posted on mumsnet and one of the sources it gave me was the same post on mumsnet with the post I thought was wrong quoted as the truth. Googling the DofE website directly showed I was right and the other poster was wrong but AI took their answer as correct.

I’d love to see your directive please!

Aren't there loads of examples of this - how many rs are in strawberry or something?

If you ask for information about sports people's medals, it literally makes up rubbish about events they didn't even compete in.

I can't believe you are shocked to be honest!

It’s only as good as the data it searches, it’s agentic capabilities and the prompts you give it. There is a clear difference between language library AI and agentic AI. Vibe coding and not. I use the Manus.ai app.

this is hilarious 😆

Asked Google:
How many times a day is AI wrong?
Answer:
There's no single daily error count for all AI, but a February 2025 BBC study found 51% of answers from public AI assistants to news questions had significant issues, including factual errors or altered quotes, and a March 2025 Mashable article reported a separate study showing AI search engines got answers wrong as much as 60% of the time. The actual rate depends on the specific AI model, its training data, and the type of task it's performing.

You know that thing where you start a sentence on your phone and then take every suggestion from the predictive text to create a nonsense sentence? What large language models are doing is basically a more sophisticated version of that based on the information on the internet. LLM AI doesn't look up answers, it tries to guess what an answer to your question might look like (and of course it doesn't understand the question either!)

Don't ever use those AI things that appear at the top of Google searches. They are ALWAYS completely wrong!

The FBk suggestions aren't just inaccurate, they are often misleading, irrelevant, even dangerous ("diet tips for anorexics" anyone ?)

I now distrust every single video that Facebook tries to feed me, I assume they are 100% faked too.

I'm holding breath for when I encounter a YouTube "how to" video that turns out to be AI-generated and has nothing to do with safely or competently doing some task I need to do, instead the instructions cause me to make some expensive or important error. Ditto with "unboxing videos" or "product reviews" online : would assume these are all manufacturer- or scammer-produced fakes nowadays, too. Aren't most the testimonial posts in comments with Facebook ads, all faked?

I moderate Facebook groups & see a lot of Ai-generated profiles. The best FBk groups have cultural norms that the AI-bots instantly violate making the AI-profiles esp. obvious.

Basically if it wasn't online before 2021 then best to assume it's AI-generated nonsense, and even before 2021, reasonable to suspect it's also rubbish online info.

An (elderly) lady I volunteer for was asking me about Frive yesterday: "Do you know about it? Why is it necessary?" Frive seems to just be a ready-meal service. I tried to explain "They are just trying to sell you something". She thought Frive was some useful new vitamin or wonder food.

Of course they aren't always wrong, that would mean that either all the information available on the internet is wrong or that lLMs ignore all the correct information and always pick the nonsense

Which of those do you think it is?

The only way I figured out what chat GPT is was by Googling an explanation.

It's a large language model, it works by predicting the next word in a sentence.

AI also looks a huge sets of data and makes predictions.

I think it's worrying that we might unskill people in critical thinking skills personally.

AI is more often wrong than right if I ask it serious questions, it's information is often out of date just depending how often stuff on the internet gets updated. Popular topics may be up to date, less popular probably not. And it gives the most commonly given answers to things which are not necessarily the correct ones! It has no discernment and is totally open to manipulation.

I actually said to ChatGPT just this morning "Jesus, if AI is the future then god help us"