News Microsoft Brings ChatGPT AI to Bing and Edge

bigdragon

Distinguished
Oct 19, 2011
1,125
577
20,160
Great! I love it when search engines search for what they think I meant to type instead of what I actually typed. Clearly I mistyped TPM (Trusted Platform Module) and intended to type TPMS (Tire Pressure Monitoring System). /s

I still think it's way too early to be rolling these AI features out. However, I'm sure some clickbait news blogs will love this. They can book work trips via the AI and then write about itinerary mistakes. Maybe have the AI write some department-wide emails that are mildly embarrassing but good enough for several articles. I think AI is going to be great in 4-8 years -- and a technology that forces us to re-evaluate the relationship between work, income, and living. However, we're not there yet.
 

PlaneInTheSky

Commendable
BANNED
Oct 3, 2022
556
762
1,760
This is of course not "AI".

ChatGPT is simply able to answer in contextual human-like speech because it is piecing together sentences it sourced online.

This is also not new. Google has been adding answers to contextual questions for years. "How tall is the Eiffel Tower ?" is something Google has been able to answer for over a decade now.

The difference is that Google shows where they found this information. And the contextual data used so far has been limited to such an extent that it is still manageable for Google.

ChatGPT is just taking contextual answers to an extreme, and slicing and combining information from all over.

ChatGPT does not disclose where it found this information, how it found this information, and the date of this information.

A lot of data from ChatGPT is flat out wrong, outdated, or both.
And while asking ChatGPT to write a poem is no big deal, people will be asking medical information, financial information, etc... and that's going to go horribly wrong.
 
Last edited:
  • Like
Reactions: Why_Me
D

Deleted member 14196

Guest
What a nice way to ruin a search engine. Oh wait this is bing. It needs all the help it can get

bing = bung

ANTI Intelligence more like it at this point. Here they are laying off all these high tech people and their investing money in this stuff. What a joke.
 

wbfox

Distinguished
Jul 27, 2013
81
40
18,570
Copy one persons work and you are a plagerist. Copy the work of 1,000 and you are Google but to copy the work of 1,000,000,000, well, it just means your a natural language process ai startup. -StallinRA
 

Geef

Distinguished
bing = bung

Haha if Microsoft renamed their Bing search browser to Bung or Bunghole it would get WAY more popular. Everyone would know about it just from the name. They would say during conversations instead of 'ask Google' they would say 'Ask your Bunghole app!'
 
D

Deleted member 14196

Guest
I’m envisioning Beavis playing cornHolio
 

taz-nz

Distinguished
Apr 27, 2006
41
31
18,560
I love it when search engines search for what they think I meant to type instead of what I actually typed.

I know you were joking but have tried to use Google to search for a lesser-known topic or odd brand name that's similar to an English word lately. You end up with a search that looks like ++++++""""what I'm actually looking for"""" -------"no not that" -------"or that" ------"why the hell did you think I meant that???!!!"
 
  • Like
Reactions: TJ Hooker

bit_user

Polypheme
Ambassador
have tried to use Google to search for a lesser-known topic or odd brand name that's similar to an English word lately.
Are you signed into your google profile? I know this is going to sound like I'm shilling for google, but one thing their profiling does is try to learn enough about you that it can figure out which of those things you meant.

Try designing a search engine for all languages and spelling abilities, and you'll quickly run into the same problem - that a lot of people can't spell - and that makes it even harder to tell exactly what they want. Especially absent context.

Having said that, I don't do things like that to "help" google, because my search results are usually good enough. I'm just saying that, if this is a real problem for you, you might find you're not seeing the best performance their tools can offer.
 

bit_user

Polypheme
Ambassador
This is of course not "AI".

ChatGPT is simply able to answer in contextual human-like speech because it is piecing together sentences it sourced online.
It doesn't. It has a knowledge base that it learned from the text it reads, and it synthesizes text from that knowledge base. Kind of like a human would do, if you asked them trivia questions after they'd read all the books in a library. Some clues to watch for are things like how it can adjust its speaking style - the fact that it can present the same information in different styles is telling you it's not just copy-and-pasting text.

What amuses me about your statement is how you seem to think "piecing together sentences" is even an easy problem, if you want the text to read and flow in a logical and consistent fashion.

ChatGPT does not disclose where it found this information, how it found this information, and the date of this information.
I'm guessing you didn't read the article, because they actually addressed this point.

A lot of data from ChatGPT is flat out wrong, outdated, or both.
That's a bit like complaining that the Wright Brothers' first plane couldn't take passengers or do long-haul flights. The technology is moving fast, and ChatGPT was just a tech demo that's already obsolete.

And while asking ChatGPT to write a poem is no big deal, people will be asking medical information, financial information, etc... and that's going to go horribly wrong.
As mentioned in the last article's comments, this is a solvable problem. They could include automatic disclaimers about those topics, or even have it flat-out refuse to answer.
 

bit_user

Polypheme
Ambassador
Copy one persons work and you are a plagerist. Copy the work of 1,000 and you are Google but to copy the work of 1,000,000,000, well, it just means your a natural language process ai startup.
Copyright laws protect the exact form of a work, not the knowledge behind it. I think you highlight an interesting problem, which is that we probably have too few tools for IP protection.

Anyway, most of what we all know are things we learned from somewhere else. Even the ideas you think are original probably aren't as original as you believe. Anyway, what ChatGPT is doing is very much analogous to the way we all read stuff online and repeat in other contexts (usually without attribution). The difference is in its scale and quite how good it is at remembering certain things.

BTW, I'd encourage you to read the article, since Microsoft seems to be addressing the attribution problem.
 
Last edited: