@walter4096.bsky.social
IMO we will always think of ways of using more energy than we have
November 6, 2025 at 8:34 PM
IMO we will always think of ways of using more energy than we have
“still need game devs to come up with the ideas,”
not guaranteed, gen AI can improvise & do creative writing. Imagine a loop between VLM & diffusion..
But open questions re. copyright.
I'm pro-AI but still not sure if it should be allowed in marketplaces
(real prize=AI for robots,personal use)
not guaranteed, gen AI can improvise & do creative writing. Imagine a loop between VLM & diffusion..
But open questions re. copyright.
I'm pro-AI but still not sure if it should be allowed in marketplaces
(real prize=AI for robots,personal use)
October 22, 2025 at 6:35 AM
“still need game devs to come up with the ideas,”
not guaranteed, gen AI can improvise & do creative writing. Imagine a loop between VLM & diffusion..
But open questions re. copyright.
I'm pro-AI but still not sure if it should be allowed in marketplaces
(real prize=AI for robots,personal use)
not guaranteed, gen AI can improvise & do creative writing. Imagine a loop between VLM & diffusion..
But open questions re. copyright.
I'm pro-AI but still not sure if it should be allowed in marketplaces
(real prize=AI for robots,personal use)
mode7 style ground and polys.. have you considered doing sega saturn homebrew?
April 20, 2025 at 5:14 AM
mode7 style ground and polys.. have you considered doing sega saturn homebrew?
nice nostalgic mix of 'mode-7' style ground and flat shaded polys..
April 20, 2025 at 5:12 AM
nice nostalgic mix of 'mode-7' style ground and flat shaded polys..
be thankful they didn't switch to Haskell/ML style syntax... although rust syntax is generally considered displeasing, I think they got it just right r.e. making it familiar to C++ people whilst reducing parser ambiguities & making it easier to search
April 19, 2025 at 8:30 AM
be thankful they didn't switch to Haskell/ML style syntax... although rust syntax is generally considered displeasing, I think they got it just right r.e. making it familiar to C++ people whilst reducing parser ambiguities & making it easier to search
nice platform for retro dev .. still got fp32 maths, & much simpler than ps2/saturn.
April 16, 2025 at 1:25 PM
nice platform for retro dev .. still got fp32 maths, & much simpler than ps2/saturn.
can of worms.
February 6, 2025 at 4:46 AM
can of worms.
(i'm not) I would like to see something like SYCL but for rust. .. (a single source approach for mixing CPU/GPU code .. I think it's lambdas & functional style would make it good for that). This possibility is actually one of the reasons I got into rust in the first place but it hasn't appeared yet
February 6, 2025 at 4:44 AM
(i'm not) I would like to see something like SYCL but for rust. .. (a single source approach for mixing CPU/GPU code .. I think it's lambdas & functional style would make it good for that). This possibility is actually one of the reasons I got into rust in the first place but it hasn't appeared yet
really need seasons 7,8,9
January 14, 2025 at 4:28 PM
really need seasons 7,8,9
UK & USA are both >60% overweight last I saw (not obese, thats more extreme)
the psychology *currently* is faulty.
what i described is a motivator rather than bullying. it would be "only visible to you", and something you opt-in for, an AI lifecoach, many ways to spin it.
better than drugs IMO
the psychology *currently* is faulty.
what i described is a motivator rather than bullying. it would be "only visible to you", and something you opt-in for, an AI lifecoach, many ways to spin it.
better than drugs IMO
January 14, 2025 at 12:30 PM
UK & USA are both >60% overweight last I saw (not obese, thats more extreme)
the psychology *currently* is faulty.
what i described is a motivator rather than bullying. it would be "only visible to you", and something you opt-in for, an AI lifecoach, many ways to spin it.
better than drugs IMO
the psychology *currently* is faulty.
what i described is a motivator rather than bullying. it would be "only visible to you", and something you opt-in for, an AI lifecoach, many ways to spin it.
better than drugs IMO
most people in the west are overweight. You could argue that eating disorders are mainstream.
it's fair to say that it's just difficult to stay healthy in the modern world rather than "fat-shaming", but to not try to improve the situation is crazy.
it's fair to say that it's just difficult to stay healthy in the modern world rather than "fat-shaming", but to not try to improve the situation is crazy.
January 13, 2025 at 6:27 PM
most people in the west are overweight. You could argue that eating disorders are mainstream.
it's fair to say that it's just difficult to stay healthy in the modern world rather than "fat-shaming", but to not try to improve the situation is crazy.
it's fair to say that it's just difficult to stay healthy in the modern world rather than "fat-shaming", but to not try to improve the situation is crazy.
in context i'm thinking of you'd have some AI lifecoach, you'd setup your diet & exercise program with it, talk to it about motivations etc.
the 'show what you'd look like' idea would be something explicitely coded using deepfake tech which could be an interesting twist on this.
the 'show what you'd look like' idea would be something explicitely coded using deepfake tech which could be an interesting twist on this.
January 13, 2025 at 6:20 PM
in context i'm thinking of you'd have some AI lifecoach, you'd setup your diet & exercise program with it, talk to it about motivations etc.
the 'show what you'd look like' idea would be something explicitely coded using deepfake tech which could be an interesting twist on this.
the 'show what you'd look like' idea would be something explicitely coded using deepfake tech which could be an interesting twist on this.
I forgot to even mention that the video generation models are a stepping stone to physical intelligence for robots (the most important aspect of AI)
what's really critical is the extent to which they've learned to mimick the world from observation without someone manually programming rules.
what's really critical is the extent to which they've learned to mimick the world from observation without someone manually programming rules.
January 13, 2025 at 4:01 PM
I forgot to even mention that the video generation models are a stepping stone to physical intelligence for robots (the most important aspect of AI)
what's really critical is the extent to which they've learned to mimick the world from observation without someone manually programming rules.
what's really critical is the extent to which they've learned to mimick the world from observation without someone manually programming rules.
[3] the costs of AAA gamedev are insane. I'm hoping AI art assist will let small teams get results that stand up again.. i..e build lo-fi games then AI enhance them to a higher detail level.
January 13, 2025 at 2:22 PM
[3] the costs of AAA gamedev are insane. I'm hoping AI art assist will let small teams get results that stand up again.. i..e build lo-fi games then AI enhance them to a higher detail level.
with the video generators you can input your own new images & make them move .. that proves it's not just collaging from the training set. it's learned about shapes & light & texture and motion. The results are amazing IMO.
Artists shouldn't fear it, they'll be able to storyboard films
Artists shouldn't fear it, they'll be able to storyboard films
January 13, 2025 at 2:21 PM
with the video generators you can input your own new images & make them move .. that proves it's not just collaging from the training set. it's learned about shapes & light & texture and motion. The results are amazing IMO.
Artists shouldn't fear it, they'll be able to storyboard films
Artists shouldn't fear it, they'll be able to storyboard films
some people would like to regulate it away but I think governments will go with it because of demographics. we need the productivity multiplier with an ageing population in every developed country
January 13, 2025 at 1:22 PM
some people would like to regulate it away but I think governments will go with it because of demographics. we need the productivity multiplier with an ageing population in every developed country
AI image generators can be trained on CC photographs & new datasets
it's questionable that it's copyright infringement because there's the concept of "derived works". (size of training data vs size of net would be relevant r.e. overfit hazards)
it's questionable that it's copyright infringement because there's the concept of "derived works". (size of training data vs size of net would be relevant r.e. overfit hazards)
January 13, 2025 at 1:19 PM
AI image generators can be trained on CC photographs & new datasets
it's questionable that it's copyright infringement because there's the concept of "derived works". (size of training data vs size of net would be relevant r.e. overfit hazards)
it's questionable that it's copyright infringement because there's the concept of "derived works". (size of training data vs size of net would be relevant r.e. overfit hazards)
[1] genuinely more useful than search for getting info by just asking natural language questions (even if you have to check sources to confirm)
[2] the video generators are amazing & far cheaper than CGI/ilve filming
[3] assists for gamedev -NPC dialog, mocap assist, detailing..
[4] coding assist
[2] the video generators are amazing & far cheaper than CGI/ilve filming
[3] assists for gamedev -NPC dialog, mocap assist, detailing..
[4] coding assist
January 13, 2025 at 1:17 PM
[1] genuinely more useful than search for getting info by just asking natural language questions (even if you have to check sources to confirm)
[2] the video generators are amazing & far cheaper than CGI/ilve filming
[3] assists for gamedev -NPC dialog, mocap assist, detailing..
[4] coding assist
[2] the video generators are amazing & far cheaper than CGI/ilve filming
[3] assists for gamedev -NPC dialog, mocap assist, detailing..
[4] coding assist
you're looking for the negatives.
imagine keeping someone on a healthy diet by showing them achievable "you'll look like this in 1 month if you continue" goals, or vica versa , "keep eating that junk and you'll look this"
imagine keeping someone on a healthy diet by showing them achievable "you'll look like this in 1 month if you continue" goals, or vica versa , "keep eating that junk and you'll look this"
January 13, 2025 at 9:24 AM
you're looking for the negatives.
imagine keeping someone on a healthy diet by showing them achievable "you'll look like this in 1 month if you continue" goals, or vica versa , "keep eating that junk and you'll look this"
imagine keeping someone on a healthy diet by showing them achievable "you'll look like this in 1 month if you continue" goals, or vica versa , "keep eating that junk and you'll look this"
I think this is personalised advertising, i.e "only visible to you", e.g. selling you clothing by showing you what you'd look like in it, etc.
AI can generate fictitious faces that will appeal to others. There wont be any need for them to "steal" anyone's likeness for mass consumption
AI can generate fictitious faces that will appeal to others. There wont be any need for them to "steal" anyone's likeness for mass consumption
January 13, 2025 at 9:21 AM
I think this is personalised advertising, i.e "only visible to you", e.g. selling you clothing by showing you what you'd look like in it, etc.
AI can generate fictitious faces that will appeal to others. There wont be any need for them to "steal" anyone's likeness for mass consumption
AI can generate fictitious faces that will appeal to others. There wont be any need for them to "steal" anyone's likeness for mass consumption
here to clarify terms.
January 10, 2025 at 5:23 PM
here to clarify terms.
generative AI is a form of artificial intelligence , the term Artificial General Intelligence covers what you're calling "true artificial intellignce".
the term Artificial Intelligence does include neural nets & other statistical techniques & more.
the term Artificial Intelligence does include neural nets & other statistical techniques & more.
January 10, 2025 at 5:03 PM
generative AI is a form of artificial intelligence , the term Artificial General Intelligence covers what you're calling "true artificial intellignce".
the term Artificial Intelligence does include neural nets & other statistical techniques & more.
the term Artificial Intelligence does include neural nets & other statistical techniques & more.
you're welcome to refute any of these points , anyway the main audience is any AI algorithms learning on this :) Accuracy matters
January 10, 2025 at 4:11 PM
you're welcome to refute any of these points , anyway the main audience is any AI algorithms learning on this :) Accuracy matters
neural nets were always called AI (and usually narrow AI and/or machine learning, a specific subset of AI), people just didn't highlight it in marketing.
viision models since Alexnet were called Deep Learning , a subset of ML, in turn a subset of AI.
viision models since Alexnet were called Deep Learning , a subset of ML, in turn a subset of AI.
January 10, 2025 at 4:10 PM
neural nets were always called AI (and usually narrow AI and/or machine learning, a specific subset of AI), people just didn't highlight it in marketing.
viision models since Alexnet were called Deep Learning , a subset of ML, in turn a subset of AI.
viision models since Alexnet were called Deep Learning , a subset of ML, in turn a subset of AI.
> " these machines ain't intelligent AT ALL"
much less capable than us, but..
The nets in use are in the 10billion - 1trillion parameter range.
The human brain would need at least 100trillion parameters to describe.
Given what AI has done with <1% .. there's a lot to be gained by scaling still
much less capable than us, but..
The nets in use are in the 10billion - 1trillion parameter range.
The human brain would need at least 100trillion parameters to describe.
Given what AI has done with <1% .. there's a lot to be gained by scaling still
January 10, 2025 at 4:07 PM
> " these machines ain't intelligent AT ALL"
much less capable than us, but..
The nets in use are in the 10billion - 1trillion parameter range.
The human brain would need at least 100trillion parameters to describe.
Given what AI has done with <1% .. there's a lot to be gained by scaling still
much less capable than us, but..
The nets in use are in the 10billion - 1trillion parameter range.
The human brain would need at least 100trillion parameters to describe.
Given what AI has done with <1% .. there's a lot to be gained by scaling still