OpenAI and table tennis

Garbage! The 0X version of T05? Since when does T05 come without sponge.
What is a tensor rating? What are the units?

Hit the ball higher?

OpenAI must of scanned a TT forum full of myths.

OK, I looked into OpenAI GPT-3. It is still immature but I am sure it will get better. One of the problems is the old "garbage in, garbage out" problem. Neural nets and knowledge bases need to be fed correct data.
I blame the poor results above on bad source data and much of it comes from TT forum and manufacturers.

Anyway, I want to thank palguay for making me aware of this project. If I get access to a natural language processor that can truly have a clue then I have years of magazine articles that I have written that could be fed into the OpenAI knowledge base for my company's use and maybe my customers. I would apply this to motion control and control theory.

Lawyers, law firms, already have access to knowledge bases for searching case law. At lot of cases can be simply looked up because the same kind of case was done before. All the arguments and the results are know. However, there are new cases that don't quite fit what has happened before. Being able to search for similar cases is handy.

I don't see where OpenAI will be very handy for table tennis. I do think it will enable intelligent machines to replace more people. I am 67 and retired now. It amazes me how many people are worried about what rubber or blade to buy today and not how they are going to afford it in the future.

Don't speak to it. Don't speak to higher level intelligence. It may recognize you are inferior, and you don't know what happens then.

The only higher level intelligence you are allowed to speak to are the table tennis gods. They are benevolent, and will not destroy you. They may even feel pity and compassion on good days.

Slightly more serious:
https://bdtechtalks.com/2020/08/17/o...commercial-ai/

It's interesting to see how much the training of such huge net costs and how much energy it takes. But that is a detail...

P.S. Poor Carl, already now fighting bots. Imagine where this will evolve. How about using "good" AI against the "bad" AI, soon or later...

Originally Posted by latej

Don't speak to it. Don't speak to higher level intelligence. It may recognize you are inferior, and you don't know what happens then.

The only higher level intelligence you are allowed to speak to are the table tennis gods. They are benevolent, and will not destroy you. They may even feel pity and compassion on good days.

Slightly more serious:
https://bdtechtalks.com/2020/08/17/o...commercial-ai/

It's interesting to see how much the training of such huge net costs and how much energy it takes. But that is a detail...

P.S. Poor Carl, already now fighting bots. Imagine where this will evolve. How about using "good" AI against the "bad" AI, soon or later...

Too late but I am not bot.
https://deltamotion.com/peter/Videos...ine%20Demo.mp4
My company deals in automation and AI or optimization. I know our machines replace people.
The video shows scanning potato strips and knives that cut out the defect of potato strips.
It was hot and miserable work in the summer.

The young people on this form need to worry more about how they will compete against the robot and smart machines that other TT players.

The world is changing 50% have an IQ of below 50%. What do you do with the lower 50%.It makes no difference is you make college free to them. They will still be the lower 50%.

Originally Posted by brokenball

The young people on this form need to worry more about how they will compete against the robot and smart machines that other TT players.

:-) By that time we may also develop a kind of training suit, which will give you feedback on those parts of the body which don't move optimally. Imagine wearing such Ma Long suit.

Originally Posted by Der_Echte

BB, people can change, a few of the jokers studying may become the next Thomas Sowell... but yeah, at the end of the day, out of a population of 100%, there will be a bottom 50% even if 100% of a nation has a college diploma.

This wisdom seems to have eluded the politicians that want to make college free for everybody.

Originally Posted by latej

:-) By that time we may also develop a kind of training suit, which will give you feedback on those parts of the body which don't move optimally. Imagine wearing such Ma Long suit.

Think of the technology that would be required to make such a suit. All you want it to do is play table tennis? Where is the return on investment? I am sure the militaries of the world would have different ideas.

@Palguay, I have been thinking about it. I saw a webpage with python code that shows how to get the gpt-3 source code from git hub. Is that what you did?
The problem is training the software with good information and knowing how to tweak the algorithms.

I have a data base of text. It would be easy enough to just search for what one wants using a search engine but I would like to try this.
Something that isn't well known is that one can search websites and forums using Google search
if I type
brockenball site:tabletennisdaily.com/forum
I can find all my posts or on this forum
The problem is that I need to make all the data I want to search public so Google's web crawler can find it.

Originally Posted by brokenball

@Palguay, I have been thinking about it. I saw a webpage with python code that shows how to get the gpt-3 source code from git hub. Is that what you did?
The problem is training the software with good information and knowing how to tweak the algorithms.

I have a data base of text. It would be easy enough to just search for what one wants using a search engine but I would like to try this.
Something that isn't well known is that one can search websites and forums using Google search
if I type
brockenball site:tabletennisdaily.com/forum
I can find all my posts or on this forum
The problem is that I need to make all the data I want to search public so Google's web crawler can find it.

No I just used the default playground provided by openAI. OpenAI is already trained we cannot train it with our data, we can provide context and it seems to be pretty good at using that to provide answers. Let me know if you have any questions