Chat GPT
-
https://www.taiwannews.com.tw/news/6030380
DeepSeek Debates: Chinese Leadership On Cost, True Training Cost, Closed Model Margin Impacts
The DeepSeek Narrative Takes the World by Storm DeepSeek took the world by storm. For the last week, DeepSeek has been the only topic that anyone in the world wants to talk about. As it currently s…
-
Have been playing with Grok3. It's incredible that in around a year, XAI has not only caught up to the leading companies, but is arguably now ahead.
It's continuing to train, and they are making daily updates to features, and to regular updates to the model. Voice is coming in a few days.
Has a huge token window, is incredible quick, and the results have been epic. You can point it at API docs online and it will summarise aspects of the results, and pull in information to support it. It's like having your own researcher. Coding is fantastic, and the news summaries are great. Realtime data from X gives it a good point of difference.
They have quietly finished the 200,000 GPU cluster, and continue to expand and plan larger. The Elon superpower of building things impossibly quickly is a HUGE advantage with this. Not sure anyone can compete, next year is going to be wild as they try.
-
Have been playing with Grok3. It's incredible that in around a year, XAI has not only caught up to the leading companies, but is arguably now ahead.
It's continuing to train, and they are making daily updates to features, and to regular updates to the model. Voice is coming in a few days.
Has a huge token window, is incredible quick, and the results have been epic. You can point it at API docs online and it will summarise aspects of the results, and pull in information to support it. It's like having your own researcher. Coding is fantastic, and the news summaries are great. Realtime data from X gives it a good point of difference.
They have quietly finished the 200,000 GPU cluster, and continue to expand and plan larger. The Elon superpower of building things impossibly quickly is a HUGE advantage with this. Not sure anyone can compete, next year is going to be wild as they try.
Bastards, that's why I can't afford to upgrade my gaming rig. First crypto, then covid, now AI
-
Grok3 is waaaaay improved over Grok2, on the political/moral/philosophical front. Maybe leaning into X posts more? Actually was able to have a balanced discussion with it on the whole DEI/equity/discrimination front and it was able to easily link me to legislation and court outcomes to back up its points.
Little sample here:
So, is equity discriminatory to the individual? By design, yeah—it prioritizes group outcomes over equal shots. Whether that’s “fair” depends on if you buy the systemic fix or see it as new bias in a noble mask. What’s your take—where’s the line for you?
-
The speed is incredible, I get the answers before I can get my hand to the mouse.
As for the wokeness, it's hard to get rid of it entirely as it is trained on the internet after all. But it's much better, I don't think left leaning people will be upset about it's answers. Leftists might though, as it doesn't lie when interpreting statistics, for example.
If they can catch up this much in a year, where the fuck are they going to be in 2026?
-
Have been having an absolute ball with Grok3.
Now that I know its starting point isn't 'I have a serious head injury, how can I help?'Had it calculate for me ROI of putting in a water tank at home, moving to an electric vehicle using salary sacrifice and taking advantage of a gov rebate to convert from gas to electric heating. I still need to play a bit with the numbers but the logic looks incredibly sound.
Also had it generate a 5min children's story for my daughter complete with characters she likes - dinosaurs, fairies, bunnies. The story was actually better than half of the dross we get at the library. Only hiccup however was when I asked it to generate a cover for it it started making some cutesy thing but then for whatever reason it kinda crashed and instead generated an image of a middle aged guy with glasses and then the same dude but shirtless in a swimming pool...super odd. I got it to correct, it said there was some error mixing with another thread which was weird unless its mixing between different peoples requests..anyway when it did eventually complete the cover it was really well done.
-
Really starting to understand just how useful AI is going to be.
Unwell daughter at the moment with really mixed off and on symptoms, has both her mum and I completely flummoxed.
Summarised it all and sent it to Grok and wow, some amazing points, things to look out for and help work out what might be going on. I mean no replacement for a GP..surely..but then again in my time I've had GPs misdiagnose things, probably not a bad method to at least check their working
-
Never thought I'd get to this point but I now use AI every day. Random internet searches, cooking, medical advice, research. We have co-pilot at work, which is ok but grok is just that much faster, feels less restricted and doesn't get as stuck with dumb responses from time to time. Really keen to try Grok voice but doesn't seem to be available on Android yet, got my brother on to it today and he gets voice on Apple, hopefully its still free when it eventually gets to android. But yeah blown away after being pretty not interested for a while.
-
@Rembrandt said in Chat GPT:
Never thought I'd get to this point but I now use AI every day. Random internet searches, cooking, medical advice, research. We have co-pilot at work, which is ok but grok is just that much faster, feels less restricted and doesn't get as stuck with dumb responses from time to time. Really keen to try Grok voice but doesn't seem to be available on Android yet, got my brother on to it today and he gets voice on Apple, hopefully its still free when it eventually gets to android. But yeah blown away after being pretty not interested for a while.
x said "Grok Android app is now open for testing in Australia, Canada, India, Philippines, and Saudi Arabia." https://x.com/ablenessy/status/1886661394499862760
But isn't it just based on X data? Or more, now?
-
@nostrildamus Mostly websites when it lists sources. Seems to lean on X more when working out context for recent events. Maybe voice is a staggered rollout..even grok can't tell me when I'm getting it
-
@Rembrandt Talking about itself seems to be a weak point for Grok - it tells me it can't generate images and immediately generates images.
-
@delicatessen That's weird, although to be fair on the image front is where I had my only real error. Started creating the images I wanted before randomly creating these specific images of a middle aged bloke with glasses. Happened a couple times so I asked it about it. Apparently the image it was creating was someone with my name that is on the internet. Very odd. Clearly still bugs to work out
-
-
Had a revealing answer from co-pilot (Microsoft's ai). Work has bought a license and we are only allowed to use co-pilot for AI. I'd finished work for the day and my daughter came into the study wanting me to print out something to colour-in. Thought it would be fun to convert an image we took of hear at a princess birthday party into a black and white sketch that she could colour in. Grok does this task easily. As I was on my work laptop I'd figure I'd give co-pilot a crack. The image was of my daughter and a childrens party entertainer both dressed up as princesses. I asked for the image to be converted to a black and white sketch that we could then print and colour in. The output was 4 photos, all in colour. 2 with Ai generated women in princess outfits..2 with AI generated men in princess costumes, 2 of the men bearded..seriously wtf? I can only assume DEI programming superseding logic. Absolute horror show.
