Grok3 - wow!

[GenXguy]

Grok was about 50% accurate in some testing I did. It got some tax law questions correct, but then it failed on a couple easy tests: Grok3, the AI on X (old Twitter) is flat out the best doctor I've ever had. Stunningly Good!

I tried a previously asked it to figure out a TV show based on various clues, that I had finally steered to the correct answer as mentioned in a recent post, but used the DeepSearch feature this time, but Grok still couldn't come up with the correct answer despite having enough clues and having been schooled on it last time to get to the correct answer. Once again, I had to provide the actress's name for it to even consider the correct TV show.

 

[omni550]

I've been playing with Grok3. It gives me a LOT of Google-esque verbiage.

My latest query is trying to resolve a recurring issue with Windows 11's automatic updates that kill my laptop's keyboard and touchpad. Updating the drivers does nothing, only reinstalling a repair version of 24H2 (which takes 3 hours) seems to fix the issue. Lately this is occurring almost daily. Hopefully Grok3 finds a workable solution.

omni

 

[GenXguy]

Anyone know how it compares to Perplexity?

I have Perplexity Pro. It has more options than Grok. I can't say it's any better or worse. But unlike Grok, Perplexity Pro using the Pro Claude Sonnet 3.7 option, came up with the correct TV show quickly without ever having to provide the actress's name, which Grok was unable to do, even using the Think or DeepSearch options.

I think this is one of those YMMV.

 

[Cujet]

Are you a stockholder?

Flieger

Not in any way. Nearly 100% of my retirement is in CD's.

 

[Scuba]

I've been playing with Grok3. It gives me a LOT of Google-esque verbiage.

My latest query is trying to resolve a recurring issue with Windows 11's automatic updates that kill my laptop's keyboard and touchpad. Updating the drivers does nothing, only reinstalling a repair version of 24H2 (which takes 3 hours) seems to fix the issue. Lately this is occurring almost daily. Hopefully Grok3 finds a workable solution.

omni

Let us know if it helps.

 

[marko]

I just signed up with Grok3 based on this discussion. Played around and got some very personalized answers.

I feel like how people must have felt when they saw their first television. Sort of a "this changes everything" moment.

 

[sengsational]

I used to ask myself how many bar fights have been prevented by everyone being able to use web search. But then the content of the web changed so that you would get any answer you wanted. I'd like to think that LLM's will agree with each other more often, decreasing bar fights again.

 

[Scuba]

I quizzed Grok3 on a number of varied topics today. Got very thorough answers again. Another thing I noticed that surprised me is that the AI seems to have empathy. One of the “conversations” we had was about my MIL being in Memory Care and how her abilities and quality of life has declined. The AI asked great questions and seemed to really care about my query which related to how to help make her life better given her difficult circumstances. I haven’t used other AI options. Are they all empathetic like this one? If I were a lonely person without a good human support system, Grok3 might become my BFF.

 

[Koolau]

Thanks for the correction. I had heard that here (I think). Not the first time I got something wrong (heh, heh, this week!)

No, it doesn’t.

Yeah, I had already responded to him earlier in the thread that I just signed up at Grok.com, no X account.

Heh, heh, I think I've got it now. Thanks again.

 

[Koolau]

I quizzed Grok3 on a number of varied topics today. Got very thorough answers again. Another thing I noticed that surprised me is that the AI seems to have empathy. One of the “conversations” we had was about my MIL being in Memory Care and how her abilities and quality of life has declined. The AI asked great questions and seemed to really care about my query which related to how to help make her life better given her difficult circumstances. I haven’t used other AI options. Are they all empathetic like this one? If I were a lonely person without a good human support system, Grok3 might become my BFF.

Yeah, I guess I just want the facts. "Simulated" empathy would put me off though that's a very personal issue and YMMV.

 

[marko]

Yeah, I guess I just want the facts. "Simulated" empathy would put me off though that's a very personal issue and YMMV.

Yes but some of the replies to my specific questions came back as ".. for your age and family history i would suggest......" which while not empathetic at least customized the answer to exactly what I was asking. Kinda cool.

 

[pacergal]

Its pretty good. It has an annoying habit of pandering to you with phrases like "gee that is an interesting idea" or if you have been discussing some topic, it will try and crudely segue that topic into the new one. Something like "Just like your early retirement plan, shopping for car insurance can present several challenges".

It also makes up stuff. I think they call it hallucinating. I asked for 10 books similar in plot to the comic book, movie, Superman. Specifically give me books about a single protagonist with a special ability or power in a world where nobody else has these (so not a X-men or Harry Potter). It came up with some fairly reasonable books but gave me two books that did not actually exist. When I called it out on that, it looked again and apologized, claiming the hallucinating thing.

Wait...it 'created" books that don't exist??
What else is it "creating" in the answers to questions that don't exist?

 

[Fermion]

Wait...it 'created" books that don't exist??
What else is it "creating" in the answers to questions that don't exist?

Well, that is the $100 (or more like $180 billion) question isn't it?

 

[GenXguy]

Wait...it 'created" books that don't exist??
What else is it "creating" in the answers to questions that don't exist?

Yeah, I was trying to figure out the name of a TV show a few weeks back using ChatGPT, and it came up with some suggestions first that I said were not correct, then it came up with a TV series and listed the seasons it ran. I couldn't find any info about the series, online, and I asked ChatGPT for a reference because I couldn't find any information that it exists. And it responded, "You're absolutely right. It doesn't exist."

 

[calmloki]

Yeah, I was trying to figure out the name of a TV show a few weeks back using ChatGPT, and it came up with some suggestions first that I said were not correct, then it came up with a TV series and listed the seasons it ran. I couldn't find any info about the series, online, and I asked ChatGPT for a reference because I couldn't find any information that it exists. And it responded, "You're absolutely right. It doesn't exist."

And no one has said "Create it"? I mean, it sounds like we're on a "make it so" or "Let there be light" ledge.

 

[MichaelB]

Wait...it 'created" books that don't exist??
What else is it "creating" in the answers to questions that don't exist?

A most important question. This is why an AI system is most useful when you already know a great deal about the subject being researched. (Aside from entertainment)

 

[Sunset]

I’ve been using it quite a bit over the last week or so, mostly on estate planning stuff. It is very good. It isn’t 100% correct, and does on occasion make mistakes so you can’t rely on it 100%. .......

So true, I asked it for the winning number for the next powerball draw... Nope I didn't win

 

[joesxm3]

Grok is even more paranoid than I am. It helped me design better home network security and convinced me to BTD on a new computer and $600 worth of network firewall gear.

 

[Koolau]

Okay, so I finally tried it. I gave it a lay-up question: What movie contained an image of a supposedly haunted Frank Lloyd Wright house? It spat back the correct answer "House on Haunted Hill" with Vincent Price 1959.

So I've lost track of a friend. He was quite ill but I can't find any update info on him (no obit, etc.). But just using google, I can find out where he has lived, what w*rk he was involved in, etc. But the AI couldn't find a thing though I shared everything I knew about him. Maybe it doesn't do "people searches."

 

[Kings over Queens]

"who is koolau from early retirement dot org." (insert evil laugh)

 

[Kings over Queens]

We've got an base level AI system embedded in our work computer that our evil overlords are dying to get us to use. the more its used ,the better it gets, and we're told that higher utilization helps our multiple.

Meh.

 

[DuluthGA]

I used 4 different AI’s to try and get a rotation schedule for 8 couples to have a monthly dinner club. I gave it exact requirements and they all spit out a rotation and none met the requirements. I’d say you have this couple in the group of 2 twice (a requirement was to only have each couple in the group of two once) and it would respond - you’re correct. How’s this grouping? - And it would still be wrong. Finally did a grid layout myself

 

[Koolau]

"who is koolau from early retirement dot org." (insert evil laugh)

"I'm sorry Dave but I'm afraid I can't do that."

 

[Kings over Queens]

lolz

 

[MichaelB]

I want an AI engine that can suggest what to say to DW when she’s upset with me to calm the angry spirits and get her to smile.

Another would be what to say to a teen family member without getting an eye roll or “whatever” in response. IMO those are challenges that have real life application potential and also value.