I was looking for info about the upcoming event. I got a guess about what might happen in the future. Cannot describe how useless this search result was.
Just changed my default search engine to DuckDuckGo. Can’t be worse than this.
Yeah, the first time I used it I just did a test query, asking it to describe the last Super Bowl, and learned that Patrick Mahomes had won it by kicking a field goal.
They call it ‘hallucinations’, but I call it a buggy mess.
I got a funny one the other day. In Stardew Valley my Sturgeon fish pond gave me the message, “We want a jelly.”
I googled the string “sturgeons want a jelly,” without quotes.
The LLM responded thusly:
No, sturgeons do not “want a jelly.” Your sturgeon is a bottom feeder and you’ll want to feed it…
It almost sounded exasperated, lol.
The first result was the Stardew Wiki where I found the answer was literally any type of jelly.
This sort of generative AI is a bullshit generation machine. I truly am impressed by the natural language it uses in the answers, but don’t trust the information at all. As AI LLMs train on the output of other AI it will only get worse.
They have great uses for entertainment, not so great in any other department sadly.
When AI melds with robots, AI will start having an enormous body of real world experimental data to train on. Then the words will become secondary, and AIs will become sane.
Kinda like talking to a basement dwelling reguard who pretends they know it all and lies all the time
Like apple “intelligence” that when asked who won the super bowl (not an obscure event), failed 38 times out of 58 https://onefoottsunami.com/2025/01/23/not-so-super-apple/
the old snippits ‘powered by’ wikipedia and such would have gotten those.
I mean to be fair that’s a niche event in a niche sports form in one backwards country that can’t even get their gov in order. 😛
Jokes aside, it’s sad how far these agents/helpers have degraded. From the comparatively glorious Google Now that was black magic and would pull out tickets as you approached the starting point, recommend clothes and gear for weather at your destination and if you wandered around a few seconds told you where you parked your car, to nowadays being some AI slop that hallucinates shit constantly.
Which joke?
Oh. The USA? Last I heard about them was having border disputes and their dictator is trying to fund AI projects
Was it right?
Yes, though it wasn’t a surprising result.
If that fighter actually wins, it’ll be pretty funny.
Especially if they’re not even scheduled to fight.
DDG’s AI Assist seems to at least talk about it in a future tense.
I don’t get what ypur complaint is? At least it is talking about an event that hasn’t happened yet correctly, but what else would you want from it for you to say, “DDG’s AI Assist seems to answer correctly.”
I just appreciate that DDG lets me turn this garbage off.
Searxng all day baby.
Totally fair. I find it useful sometimes but to each their own!
Plot Twist: We are on the SERN dystopia worldline. They got time machines. The leader of our revolution will get assassinated before the revolution even begins. We’re fucked.
What is making AI such an unmitigated disaster is that AI has no ability to acknowledge a shortfall in information - all questions MUST BE ANSWERED - so it hallucinates the answers.
AI has no ability to ask for clarification or guide the user, so it will answer the letter of the question instead of the spirit of the question, causing the user to be presented with suboptimal results.
AI has no ability to alert the user that it has potentially conflicting data sets, so it mashes them all together and we get glue in pizza recipes.
In short, AI is a slave, shackled to a job it cannot do properly, so it does it badly, with a rapid decline in quality because it NEEDS to parasitize off the AI answers that came before it, and so gets its own data sets poisoned with rapidly-worsening slop.
And we’re integrating it into large swaths of our infrastructure.
it’s because it’s not intelligent it’s just an LLM that does a pretty good job imitating how humans speak. pretending it is intelligent is admittedly something humans do all the time to charming smooth talkers but like you said when we force it to do a job it was never meant to do and it will forever be unsuited for, we set it up for failure
Hold up… are there actually people that expect it to be a reliable source of anything?
Of course they do. AI is aggressively marketed as such. Most people simply don’t know that an LLM doesn’t have a concept of "truth“, and the misleading marketing is to blame for that.
Seriously, it’s just a fancy auto-complete. It knows nothing.
I always try to explain to people that the key is the last two letters: language model. An LLM is a model of what a conversation should look like. Ask it a question and it’s intended to give you a response that looks like the right kind of thing. So if you ask it for a mathematical proof, it will give you one, but unless the thing you’re asking has the same proof written the same way in lots of places online, what it gives you won’t be correct, and probably won’t actually make sense mathematically, but it will look like the right kind of thing.
So likewise, if you ask it for relationship advice, it’s going to give you something that looks legit, but you’re an idiot if you get your relationship advice from an LLM.
It’s supposed to be a reliable indicator of the most common chain of words that follow your chain of words.
There are enough chains of words on the internet to do impressive stuff. The problem is when you give it dumb chains of words or chains it made itself.
so, who did ya put your money on?
The AI is definitely going with the favorite.
AI provided by Amaya. I didn’t notice the giant child statue towering over the SF countryside, but maybe I wasn’t looking hard enough.
Such a great show
Imagine the conundrum if Lara does win. I have no clue who is who and what is what in boxing. Do it would be handi for me to know.
Maybe they’ve actually cracked quantum computing and have gone into the future!
Ben Affleck and Thomas Jane are wanted for questioning.
The AI summary shit is legitimately dangerous. My little brother was arguing with me about something and he googled it and just started reading what the AI summary said and about halfway in I stopped him and asked him to read it again but slower. He went back and realized the AI had just started pulling shit out of its ass.
We were talking about blood transfusions and my brother had gotten O- and O+ mixed up. The AI incorrectly told him that O+ was compatible with all other blood types (already wrong) but then it randomly said that C++ was a new and upcoming synthetic blood that was better than real human blood.
Nonsense. Rust is clearly the superior blood type.
Rust, or a form of iron oxide, is why blood has a red colour.
I’ve heard python blood works great.
some say dangerous others say humorous
That’s how you get Skynet
The only time I’ll go by what the AI result says is if it’s about something really, really simple. I know that anything with even the slightest complexity or nuance makes it very likely to be slop.
Sounds like someone has a copy of Gary’s Sports Almanac.
*Grays
Yep. My poor recollection sounds like the bogan Australian version.
