Grok3, the AI on X (old Twitter) is flat out the best doctor I've ever had. Stunningly Good!

[dirtbiker]

Great for patients and great for physicians but bad for payers and clinics.

This is actually changing with the advent of time-based coding. Only a few years ago, we could charge only based on number of diagnoses and medical complexity. I could spend 45 minutes with a patient discussing only their depression, and I could only bill their insurance for a low level visit because I only addressed one diagnosis. However, we are now able to bill based on either medical complexity OR time. This does allow me to spend more time with my patients and still receive appropriate compensation. It's a win-win for physicians and patients. However, it costs insurance companies more. Though, this isn't always entirely accurate. Often times when we only have 15 minutes to see a patient, we're asking them to come back every 3 months, sometimes more often, so we can cover everything. If we get 30-40 minutes per person, we generally don't need to see them as often, and in this case, we aren't costing the insurance companies more money, and then everybody wins.

I'm in an administrative position (still mostly clinical, but administrative duties as well) in my very large practice, and have been working with some of the older physicians to change how they bill if they're spending the time with their patients. Many are very apprehensive of changing how they bill after years and years of practice.

 

[ransil]

Just started playing with Grok3 , looks like i hit the free no login limit tweeking my searches about 5 deep.

"Sign up to continue with Grok 3​

Enjoy higher rate limits, history, attachments, temporary chats, and more when you sign up for free"

signed up for free, its letting me tweek as needed.

will use the others and see if they agree

 

[Koolau]

This is actually changing with the advent of time-based coding. Only a few years ago, we could charge only based on number of diagnoses and medical complexity. I could spend 45 minutes with a patient discussing only their depression, and I could only bill their insurance for a low level visit because I only addressed one diagnosis. However, we are now able to bill based on either medical complexity OR time. This does allow me to spend more time with my patients and still receive appropriate compensation. It's a win-win for physicians and patients. However, it costs insurance companies more. Though, this isn't always entirely accurate. Often times when we only have 15 minutes to see a patient, we're asking them to come back every 3 months, sometimes more often, so we can cover everything. If we get 30-40 minutes per person, we generally don't need to see them as often, and in this case, we aren't costing the insurance companies more money, and then everybody wins.

I'm in an administrative position (still mostly clinical, but administrative duties as well) in my very large practice, and have been working with some of the older physicians to change how they bill if they're spending the time with their patients. Many are very apprehensive of changing how they bill after years and years of practice.

Thanks for the information. The newer method seems much more appropriate.

I wonder if that is why my previous PCP always hustled me out as quickly as possible and even whined when I brought up more than one issue per visit - while my new PCP seems very leisurely and takes the time to answer all my questions.

I really like my new(er) PCP!

 

[badatmath]

Yeah, same here. Windows 10 desktop computer and FF. I just tried two more times, one very short question, but after submitting my question, it brings up a login. Also tried in Chrome.

I just tried again. Works perfect no login. I don't use a VPN or anything. Something about your setup is different.

It gave me a very detailed analysis of "should I take my pension now or later". I'm badatmath so IDK if it is correct but it was more or less what I expected it to say ~ waiting is better if I live beyond 74.

 

[dirtbiker]

Thanks for the information. The newer method seems much more appropriate.

I wonder if that is why my previous PCP always hustled me out as quickly as possible and even whined when I brought up more than one issue per visit - while my new PCP seems very leisurely and takes the time to answer all my questions.

I really like my new(er) PCP!

It likely is. It's really tough when you get 15 minutes for a follow-up, check-up, discuss vaccines, discuss screenings, go over labs, and discuss any concerns, and then order appropriate testing and follow-up, and making sure it is all documented in the electronic medical record, all while trying to stay on time so the next patients don't complain that we are running behind. The old model of cranking out as many invoices, err, I mean seeing as many patients, as you can in a day just isn't good medicine, even if it does fatten up paychecks...

 

[Cujet]

I just tried again. Works perfect no login. I don't use a VPN or anything. Something about your setup is different.

It gave me a very detailed analysis of "should I take my pension now or later". I'm badatmath so IDK if it is correct but it was more or less what I expected it to say ~ waiting is better if I live beyond 74.

Ask it about the effect of inflation. It was better for me to wait 5 years by raw numbers. But the $8000 addl per year (by waiting) is gobbled up by typical inflation over those 5 years, and there is no net gain if inflation is just 2.5%.

It changed my mind.

 

[badatmath]

I went back to grok and fed it my iron level blood test results and asked how much to take as my doctor wasn't very specific. . . it seemed to err on the side of conservative from the reading I've done.

Definitely fun to use but IDK that I'd use it for anything major either.

I think what I liked most is it answered what I asked instead of giving me a bunch of crap I didn't want to know. Like if i asked online which iron type/dose was best I'd get 50 answers telling me to cook with a lucky iron fish . . . okay but not what I asked kwim?

 

[GenXguy]

I just tried again. Works perfect no login. I don't use a VPN or anything. Something about your setup is different.

No VPN here in use while doing and no proxy in use. I even tried a new private window in FF, still won't give me any answers and prompts for a login. I don't have any twitter or X cookies on my computer because I never used those apps in the past.

 

[badatmath]

No VPN here in use while doing and no proxy in use. I even tried a new private window in FF, still won't give me any answers and prompts for a login. I don't have any twitter or X cookies on my computer because I never used those apps in the past.

I was using FF 135.0.1 (64 bit); windows 10.

I haven't used those apps before but your visit today will put a cookie you might want to delete. . . don't have any other ideas. I have asked it 3 questions today, none super complicated but interesting to me anyway.

Kind of funny I tried to find something else to "chat" up to test and I think I stumbled on some site NSFW that chats on certain topics. . . Like oops NOT what I wanted.

 

[GenXguy]

I was using FF 135.0.1 (64 bit); windows 10.

I haven't used those apps before but your visit today will put a cookie you might want to delete. . . don't have any other ideas.

Same versions. No cookie for me - I use an extension that automatically deletes cookies after I close the tabs unless I whitelist. I verified as well. Oh well, I'll register if I have time to mess with it.

 

[Fermion]

Grok3 started failing on me with pretty simple math. Here is EV calculation it did:

"Efficiency: Range ÷ Battery = 120 miles ÷ 98 kWh = ~1.22 kWh/mile (towing efficiency). "

 

[Cujet]

Grok3 started failing on me with pretty simple math. Here is EV calculation it did:

"Efficiency: Range ÷ Battery = 120 miles ÷ 98 kWh = ~1.22 kWh/mile (towing efficiency). "

Haha, that's EV math for ya. On a serious note, Grok3 came up with the following for me:

Range=Battery CapacityEnergy Consumption per Mile=100 kWh1.25 kWh/mile=80 miles

• • For 18650 cells (e.g., Panasonic NCR18650B or similar), the capacity loss at 0.5C–1C is typically small, around 2–5%, depending on the specific chemistry and temperature. High-performance cells (e.g., Tesla’s optimized 18650s) may have even less loss.
  • Let’s conservatively estimate a 5% capacity reduction due to the higher discharge rate and towing stress (heat, etc.).

Effective capacity = 100 kWh × (1 - 0.05) = 95 kWh.
Step 3: Adjusted Range
Using the reduced effective capacity:
Range=95 kWh1.25 kWh/mile=76 miles

 

[Fermion]

My point was it could not even do a simple division and keep the units right. The answer it gave should have been 1.22 miles/kWh.

 

[Cujet]

My point was it could not even do a simple division and keep the units right. The answer it gave should have been 1.22 miles/kWh.

I saw that and most certainly did not miss your point. That's quite a 'glitch' It got the calculations correct for me.

 

[badatmath]

I asked Grok kind of a stupid question earlier today and the answer was at best useless. It either dropped connection or got mad when I told it so. . . so it seems it is a work in progress. Or maybe better at certain topics.

 

[Fermion]

I just tried both ChatGPT and Grok with a logic question I came up with and both got it wrong I think. My question was "How many people can be served a portion from a wheel of cheese that is cut with 4 cuts. ChatGPT said 11 and Grok3 said 15. I think the correct answer is 12? Three cuts divide it into six sections then a horizontal cut divides that into 12. Where does 15 come from?

 

[stepford]

I just tried both ChatGPT and Grok with a logic question I came up with and both got it wrong I think. My question was "How many people can be served a portion from a wheel of cheese that is cut with 4 cuts. ChatGPT said 11 and Grok3 said 15. I think the correct answer is 12? Three cuts divide it into six sections then a horizontal cut divides that into 12. Where does 15 come from?

Not so easy to visualize but I think 15 is the right answer. If you take 4 intersecting planes they divide space up into 15 regions. So if space were a big wheel of cheese...

Anyway, in 2 dimensions 3 cuts get you 7 pieces not 6.

Best argument I can give is by extrapolation from lower dimensions.

A line (dimension=1) is divided into 3 regions by 2 points (dimension=0). So here the number of regions is 2^2-1 =3.

In 2 dimensions in the example pictured above the plane (dimension=2) is divided into 7 regions by 3 lines (dimension=1). So the number of regions is 2^3-1 = 7.

In both these cases an N dimensional space is divided by N+1 hyperplanes of dimension N-1 into 2^(N+1) -1 regions. So a 3 dimensional (cheese shaped) is divided by 4 planes (2-dimensional structures) into 2^(3+1) - 1 = 15 regions.

 

[landover]

With same nature's laws applicable everywhere why AI can't be smarter, it's just a question of inventing hardware and software for that?

 

[badatmath]

I think a lot of AI is designed to get our data and our money by being available to "help" us. . . As long as you don't take it too seriously I suppose it is fine.

 

[Fermion]

Not so easy to visualize but I think 15 is the right answer. If you take 4 intersecting planes they divide space up into 15 regions. So if space were a big wheel of cheese...

Anyway, in 2 dimensions 3 cuts get you 7 pieces not 6.
View attachment 54619
Best argument I can give is by extrapolation from lower dimensions.

A line (dimension=1) is divided into 3 regions by 2 points (dimension=0). So here the number of regions is 2^2-1 =3.

In 2 dimensions in the example pictured above the plane (dimension=2) is divided into 7 regions by 3 lines (dimension=1). So the number of regions is 2^3-1 = 7.

In both these cases an N dimensional space is divided by N+1 hyperplanes of dimension N-1 into 2^(N+1) -1 regions. So a 3 dimensional (cheese shaped) is divided by 4 planes (2-dimensional structures) into 2^(3+1) - 1 = 15 regions.

Thanks! Yes last night my wife explained to me the answer was 15 also (I think she is a AGI). She said imagine you took a D4 die and put it at the center of the wheel of cheese then extended all four of the faces of the die through the cheese. This would divide the cheese into 14 regions with the 15th region being the inside of the D4 die.

 

[joesxm3]

I tried this with Grok3 and asked it to draw a picture of the lines. At first I gave me photo-realistic pictures of cheese and later several conflicting line drawings.

It was acting much more stupid that a few days ago. I wonder if I had a bad update.

 

[Fermion]

I can't get Grok3 to draw any images, how are you doing that?

 

[aja8888]

I think a lot of AI is designed to get our data and our money by being available to "help" us. . . As long as you don't take it too seriously I suppose it is fine.

Everything these days is designed to get your money!

 

[joesxm3]

I just ask it to generate an image. Like "can you generate an image of a unicorn."

I am using it on X from a $9/month account so I don't know if I am getting a better version than you.

If you ask for an image it should tell you if it cannot make one or you can ask why.

One thing is it takes a while (20 seconds) and then the image fills in from the top. I am on an ethernet PC so maybe you are on a slower connection?

 

[Fermion]

Ok, yes I see, I got it to generate images. It is massively failing on my request though. I first asked it to generate an image of a sailboat with a rounded stern, commonly called a canoe stern, that has a outboard motor mounted to that stern. It generated 4 images of sailboats with the more common traditional flat stern and a outboard on that. I then said just show me some pictures of sailboats with canoe sterns and it generated 4 images of sailboats again with flat sterns. Not super impressed here as I can just go to google images and request canoe stern sailboat and get this immediately: