When AI Gets It Wrong!
AI problem,the AI problem,Bing,Bing bot,ChatGPT,MKBHD,artificial intelligence,new bing,Google chat bot,google bard,AI truth problem,problem with AI
AI and Bing and ChatGPT are out of control… this couldn’t have gone any other way
The WAN show:
Unhinged Bing stories (thread):
MKBHD Merch:
Tech I’m using right now:
Playlist of MKBHD Intro music:
~
#Wrong
Google Assistant 🤔
google search is spewing ads , like half of the page already which you dont need
Spot on review
Does it draw info from the internet? In that case, it will get some things wrong, because people who publish lists like ‘best games in 2023’ or ‘best horror movies’ are all subjective and it can differ from person to person. If it draws info from websites like that, then yea, it’s going to get things factually wrong. Information even from companies can be taken with doubt, like camera specs and what not (luckily they can’t lie because of legal reasons, but then don’t ask a bot for the answer, search it up from the company or from product sites).
Who’s to say what you read on the internet is even correct? I mean, you might search it up and do your own research, but the top answer might be some bullshit that you believe. Never trust anything on the internet, which is a common saying, so if ChatGPT gets stuff from the internet, then don’t trust what it says. The only way around this is watching creators you can trust or who you know have recommended good products that you have used, or review channels that play games and movies and given opinions you agree with. You can only trust people that you know don’t give out bull shit answers because you’ve followed them and agreed with them because you too have bought products they reviewed and realised that you have similar opinions to them, which is why you can trust them in the first place.
I think this will get fixed .
Its new product , give it some time this might be next generation leap
Bing is gonna change the way students do their assignments
Fundamentally people at some point forgot you had to take with a grain of salt almost anything you are told from any source… These new tools don’t change that. Even if it’s efficiently serving you content from somewhere on the internet, someone wrote that page or comment. Someone who obviously could be wrong. So why would it be foolproof? Even if it checks it’s work and shows you the source there isn’t an internet wide universal fact check. And that’s even before accounting for the errors the machine learning inherently introduces.
But i think it’s going to be much more obvious when the full deepfakes start dropping vs these chat windows.
Bing, goes rogue
Isn’t regular search engines babk in the early days of the internet (yahoo, excite, altavista etc) suffered similar problems? Wrong and irrelevant search results. We get by. The world did not descend into chaos because of it. Why would it be different now?
Great Topic. Great Take. Thanks Marques
is it really that much of a surprise when Microsoft just in bulk released it to a large amount of Bing users? It’s not like someone needs a credit card for a monthly subscription to use this…
We are not looking at a problem with chatGPT so much as a basic flaw with the web. People post things. If you buy into everything you read your in trouble. ChatGPT has even more trouble distinguishing good from bad information. We should not be surprised. With time and money an organization can change what the internet tells you at the top search level. Some of what people think chatGPT is good at would be better done with an expert system or at least a verified set of sources.
I was watching the wan show at the same Time as this and when the clip was show I was so confused
If we now have a large group of people who believe things like Qanon what do you think will happen when a slightly plausible chat AI starts feeding people it’s half assed gleanings from the internet.
I was waiting to see you put LTT backpacks in the back of a Tesla
As a medical professional and faculty, I have been warning my students about blindly using ChatGPT, after I noticed how excited they appear to be with it. I agree that it seems to be helpful when your search is about something you know a bit about. I use the paid version and, for instance I asked it about a new medication for diabetes called Mounjaro and what I got was a long expose about Kilimanjaro (the highest mountain in Africa ). Then I tried Tirzepatide which is the same drug, generic name. The answer was half true. ChatGPT reported it as a GLP‑1 agonist while it is a dual GLP‑1 and GIP agonist making it a first-in-class med able to activate both the GLP‑1 and GIP receptors. I give as much feedback as I can when I notice gross errors. Use it at your own risk!
As for the question about whether schools will adopt this, my school has held public meetings about it and explicitly said that it is allowed for a variety of subjects. They are treating it like how schools treat calculators currently, which I think is really cool
11:27 I literally did that a few times
Bing is primed to cater to the highest bidder.
As long as Chatgpt doesn’t go illuminate/anti vaxx/ flat earth/ Jan 6th domestic terrorist… IM GOOD.
Ngl releasing this half baked chatGPT add-on is the single greatest marketing move MS has done for Bing yet.
He looks like that guy from linus tech tips
Great points. Microsoft is still gonna make absolute BANK off of this though
3:11 “It’s not forming a sentence like humans do“
How sure about that are you *actually*?
Ngl, this feels like an L take cause he is not considering the huge speed these models are being created/improved on. Like the whole human touch thing just seems like it won’t be needed in a few months. Might be wrong though
Why would it recommend Rivian when it is not available worldwide?
Every media outlet: “While some say sticking a bomb in your anus and detonating it can be dangerous, here are some fun ways people are using it.”
It’s not wrong SOMETIMES. It’s unreliable ALWAYS.
If I were to ask these chatbots a question, say about the ocean, after it gives me the answers, I would then ask it if all the answers it gave me line up with information from (said reputable scientific institution). It seems asking it about something without fact checking the answers with a reputable source isn’t a good idea at this point in the evolution of this technology.
AI is viewed positively as a technological leap. More people would believe its “single result/answer” as opposed to spending time comparing a list of results from the existing search engines. Isn’t AI going to amplify the existing biases of whoever has the biggest megaphones. It will be harder to debunk misinformation.
I have a Google Nest v2 at home and also use the Google Assistant on my phone daily and gotta say that even that technology doesn’t seem to get it right yet.
I say, “OK Google, timer on 5 minutes” and it answers that there is no device connected but I can do so in the Google Home app.
Or “How long to get to the next flower shop” and it tells me something about a shop called “Next”. If I ask about the nearest flower shop it mentions one way too far away.
I also asked “Is (restaurant) open right now?” and it said, “Yes, (restaurant) is open today from 8 am to 11 pm” but I asked outside the business hours. So it lied.
And even seemingly simple things when I listen to music and say, “Skip the next two songs” it can always just skip to the next one. Even “Start playing my Chill playlist on shuffle” doesn’t work.
And these are just a few examples. Aside from small but not less important bugs where it just doesn’t even understand simple questions like “What time is it?” or starts ignoring you, I feel like there’s a major issue with the “intelligent systems” they are missing a real quality of life impact. The user still has to know how to use the device and give the right input to make it work. I wonder if these inputs could be increased with simple updates and if it would be that “easy”, why aren’t they investing in them anymore?
Nest was founded in 2010, Google started investing in 2011 and finally acquired Nest in 2018. I thought the technology would be way more stable, easier to use, and “intelligent” after that time.
Does anyone have some better insight on why that isn’t the case?
Wow. The dramatic fellows in the beginning.… Hard pass.
I hope this pushes google to improve their services
They have stagnated for long time like Chrome is still RAM hungry, Maps haven’t improved in recent years, search results are somewhat unrelated when asked exact question, their Office software is still in initial stages
I can get behind the electric car question. The I Pace vs. Rivian. The Jaguar has enough years behind it to be a proven, great and reliable option as one of the best. Rivian, is relatively new. It hasn’t had enough market time to be “proven” one of the best (even though it may be).