ChatGPT was terrible at designing marathon plans in the article, but you'd think that it would just copy and paste any one of a thousand workable plans from the internet. Or for copyright issues, copy and paste with some word and formatting changes.
This was my favorite section from your sub-4 minute mile training:
"Once you’ve built up your speed and strength, you can start to incorporate longer tempo runs into your training. Aim for a 10-15 minute run at your goal mile pace."
Sounds like Gerry Lindgren. Why not knock out some world records during your tempo run?
As someone who works in tech and has played around with ChatGPT It is proficient and accurate as the information on the internet... which in general is pretty terribly inaccurate.
The one thing that isn't very inaccurate on the internet is how to code small reusable pieces of code. This is what ChatGPT is fairly good at. Think about it almost all other aspects of knowledge on the internet is the loudest voices not the most accurate or even search influencing it in a weird way.. think Googling any sort of symptom and you having cancer is some of the top results... no matter what human aliment you Google.
Because things around tech and coding are the most accurate information on the internet, and that is used heavily by the people who create AI/ML systems, they think the rest of the worlds subjects of knowledge are the same accuracy on the internet so that IA for subjects other than tech are going to be pretty bad if that system pulls the information from the internet. It will take a while to create a safe enough AI system that can give out useful information for other subjects.
Good software engineers are good at replacing their own jobs especially the boring repetitive work. That is what AI systems are initially going to be good at (what ChatGPT is decent at). It'll take a while for software engineers working with AI to get AI good at other subjects.
How long is a while... IDK. Even if ChatGPT creates the perfect training plan it isn't a motivator.. without motivation training plans are essentially worthless.
It could do that and that would be easier and fairly standard Information Retrieval, but the whole point is that that is not what it’s doing, rather it is actually generating a plan from word and sentence patterns and some higher level “concepts” it has learned. We tend to think of the latter as intelligence and the former as a more mechanical activity.
The absolute, optimized training regime could be developed but it would require the data collected by Garmin or Strava.
Imagine having access to all the training data from thousands of competitive runners stored away in a database...
Data, wrangled up for: mileage leading up to fast race, workouts ran 3-4 months prior, what workouts they did in the days leading up, when during the day did they run, what days did they take off, how slow or fast did they run on recovery days, what races they ran prior, heart rate, hours of sleeping - optimized-formula for these kinds of questions for all, say, sub 4:10 miles... clustered for various runner types (age, speed/endurance-types)... it would be incredible.
The trove of training data and results they have must be amazing.
How much would you pay for a truly data-customized training plan that could get you from 4:16 to 4:09..? 4:04? 3:59? ----$1,000?, $5,000?, $10,000??
People/coaches come from one of two places - experience-based and science-based. Very, very few are truly data-based and the ones there are don't actually have enough data (not enough athletes, not enough quality tracking(actual numbers), poor statistical inferences, etc).
Coming from a data/math/stats/programming background myself, I'd love to put some code to a Strava/Garmin database and see what [b]actually works[b] on a day-to-day, week-to-week, month-to-month basis.
Garmin or Strava: reach out: thekrispykremlin at gmail
The point is that it is demonstrating that it lacks the least ability to learn from the net, so that it would be far better off by information retrieval.
Yes, it is far worse than what you read here. Does anyone here recommend running 2 1/2 to 3 3/4 miles at your goal mile pace, a pace their workouts has not in the least prepared them to do even for one mile?
Of course I recommend that. Everyone knows you should be training harder and longer than you race so that you arrive fresh and strong after lowering the mileage in the weeks prior to the big day. ChatGPT can quote me on that (which it is doing for all the BS it creates, just copies what it sees on the Internet with some grammar changes to obfuscate where it came from)
I kept pestering it and followed all its social media accounts and kept "accidentally" bumping into it, and it finally gave in and responded. So joke's on you.
The news has finally trickled down to the mainstream public, so now even your senile great grandmother is signing up. Servers are overwhelmed.
And honestly, I don't blame her. If there is a super intelligent AI giving out advice, anyone NOT taking advantage of the singularity will get left in the dust and go extinct. So, everyone, including your senile great great grandmother, are signing up lickity split pronto.
The news has finally trickled down to the mainstream public, so now even your senile great grandmother is signing up. Servers are overwhelmed.
And honestly, I don't blame her. If there is a super intelligent AI giving out advice, anyone NOT taking advantage of the singularity will get left in the dust and go extinct. So, everyone, including your senile great great grandmother, are signing up lickity split pronto.
You think any humans will survive this? Only the ones not involved have a chance. I have seen the movies.
ChatGPT is ignoring you like a girl not interested in talking to you.
I kept pestering it and followed all its social media accounts and kept "accidentally" bumping into it, and it finally gave in and responded. So joke's on you.
Now it's just going to tell you made up things you want to hear so you'll leave it alone and go away thinking all smug and highly of yourself.