5 ms·
OpenAI changed its plans and won’t train on customer data, Sam Altman says
- gumballindie 3y agoA sensible decision. What needs to be addressed next is open source licensing. Goes without saying that ai is the most important breakthrough since the industrial revolution. But it needs to be complementary. Every class of technology we have built so far has been made as an extension of us. Ai is not meant to replace us. It should a become part of us.
- zamnos 3y agoThere was also that Internet thing. It's pretty useful on occasion.
- gumballindie 3y agoWell some of it is. But i think the industrial revolution is what enabled a wide range of breakthroughs thus being more impactful than the internet, not that the internet is not significant. Hmm now that i think of it i cant make up my mind which is more impactful. Either way ai with the right approach is among the two. Ideally if not led by crypto bros rebranded as ai experts.
- jgalt212 3y agoThis is very limiting in that (away from web entries) there will be no training data that's not publicly available. Using ChaptGPT for domain-specific tasks may never reach the same level of efficacy where one's problem and vocabulary looks very similar to 1000s of examples on github.
- dtagames 3y agoBig caveat here -- That's if you use the API. The web interface is still open season.
- reaperman 3y agoUnless you turn it off in the settings, which is now an option. Settings -> Data Controls -> Chat History & Training (There are only 2 settings in "Settings"...this and Dark Mode)
- catchnear4321 3y agoNote that this is tied to the history functionality. You are not exclusively toggling a preference on how your data is used by OpenAI.
- dtagames 3y agoYes, giving up your history is quite a cost. One of the main benefits of the web interface is the ability to go back and continue an existing conversation.
- arthurcolle 3y agoSo just use chatbot-ui
- mycall 3y agoAzure OpenAI Services won't mind this decision since that is their bread-and-butter on this topic.