Danilo Campos 🇵🇷
This, to me, is the risk of sitting out of the discussion of #LLMs et al except for taking cheap pot shots because we don't like the values and aesthetics of the people who are capitalizing on them.
Shaping the infrastructure and economics of the system can't be ceded entirely to people with a stripmining, privacy-optional mentality.
Danilo Campos 🇵🇷
Nevertheless, writing endless checks to cloud providers doesn't seem like the way forward.
If you want the leverage of #LLMs, but don't want that much of your everyday problem solving being visible to a third party, that's entirely reasonable, as is having real transparency into the processes that create your model.
The real revolution for these technologies will come from personal ownership and privacy, but the technology to support this will need loads of actual work. Will it happen?
Danilo Campos 🇵🇷
So, about this Rabbit. I don't want to, myself, fall into the trap of cheap #AI dismissiveness. We need to be able to articulate what "good" looks like, or the bad wins by default because it's still shipping
What I do not like about this is how it goes to all the trouble of having dedicated, proprietary hardware, but... just calls out to OpenAI anyway
There are technological limitations. Getting good performance from local models isn't an overnight thing.
https://www.theverge.com/2024/1/9/24030667/rabbit-r1-ai-action-model-price-release-date
Danilo Campos 🇵🇷
Okay, now this shit I can enjoy:
Sony making a car the same way they made a Palm OS PDA: gaudy color screens, flashy trim and ecosystem integration no one asked for
Danilo Campos 🇵🇷
Good lord.
Here’s another one. It’s an “#AI voice recorder” that’s just… sending your shit to OpenAI again
You don’t need dedicated hardware for this! Where are people getting the VC to make all this junk?
It’s hardware with an app that uploads to the cloud. The phone already has hardware that can do this. CES is going to give me an ulcer this year.
Danilo Campos 🇵🇷
So this Rabbit R1 thing strikes me as the worst of all worlds
It’s going to generate proprietary e-waste while fully delegating its AI functions to a cloud service, with a dubious security model
It seems needlessly wasteful, even as it’s clear someone is going to sell exactly these functions, someday
https://www.theverge.com/2024/1/9/24030667/rabbit-r1-ai-action-model-price-release-date
Danilo Campos 🇵🇷
@mattly yeah, this is definitely the vibe I get too
It's simply more complicated than swapping the field contents like a mad libs sheet!
William Pietri
@mattly @danilo Entirely fair point. But I'm also very over the technology-led hype cycle. I think there are a lot of people who were brain poisoned by living through two legitimate tech revolutions, the Internet and mobile computing. They keep expecting the next one to come along any minute now without considering that it might be 50 or 100 years.
So many hypesters are showing up without a measurable human upside, let alone accounting for the downsides. At this point I reserve the right to spray them with my garden hose.
GNU social JP is a social network, courtesy of GNU social JP管理人. It runs on GNU social, version 2.0.2-dev, available under the GNU Affero General Public License.
All GNU social JP content and data are available under the Creative Commons Attribution 3.0 license.