AI ethics good and bad thread
There are some amazing uses of AI happening right now sitting right alongside some horrifying use cases driven by corporate greed. We aren't talking enough about this. I want to praise scientists for pushing the boundaries of what computers can do while also shaking a finger at companies that clearly see AI as a buzzword while doing real harm without penalty.
I'll alternate between some good examples and some bad examples, perhaps some trends will become clear.
AI ethics good and bad thread
Good: Unreal Engine 5 phenomenally demonstrates integration of AI in an intentional way. Their game engine introduced entirely new or reimagined tools like Nanite, Lumen and more. These features are only possible in real-time computer graphics thanks to AI image upscaling and DLSS frame generation. They're making the impossible, possible, years ahead of schedule.
All this, and ethically it's safe, it is sandboxed, pure input and output without user data on the line.
AI ethics good and bad thread
Bad: There's competition in the space of live transcription, note taking, and screen recording.
Zoom natively can use AI to offer live captions while recording. They do prompt the user that recording has started, allowing an opt-out which is appreciated. Though it is unclear how the tech works and if anything recorded is shared with third parties.
AI ethics good and bad thread
Bad continued: Worse is Fireflies AI, who's website hides in an FAQ that it is the meeting host's fault if they don't declare that the meeting is being recorded. No opt-out is offered. Very opaque technology and data privacy practices.
Any meeting host can invite this "automatic note taking tool" and send a third party company everyone's faces, voices, screen recordings with the explicit intent of letting it be analyzed by software.
AI ethics good and bad thread
Bad: All of this, and in the end the best we can hope for is #Craptions
https://www.rikkipoynter.com/
It's possibly even worse to have inaccurate captions than no captions at all. These tools often get words, grammar, and punctuation wrong. They don't understand acronyms or the names of products. They will misspell someone's name or not understand neopronouns. Sometimes they even attribute what was said to the wrong person. This hurts everyone consuming the transcript.
AI ethics good and bad thread
Good: Adobe Firefly... I can't believe I'm about to congratulate them but, here it goes! They only train their image generation content based on opted-in content they explicitly have permission and the rights too.
People have critiqued that Adobe is falling behind on AI. And it's because they are doing it ethically, the right way, get over it lol. I applaud them.
AI ethics good and bad thread
Bad: In contrast, Microsoft finished an AI product and then immediately fired their entire AI ethics team.
Strangely, I think it is interesting they even had a team to begin with. Lots of companies don't even think AI ethics is necessary, when in my opinion it is paramount. But ethics doesn't make money I guess...
AI ethics good and bad thread
Want to learn more about AI ethics? I recommend Robert Miles on YouTube https://youtube.com/@RobertMilesAI
AI ethics good and bad thread
A video about what the heck I am talking about around Unreal's frame generation, upscaling, etc
https://youtu.be/RaY_FDaydoQ