7 ms·
Mircosoft's implementation of ChatGPT in Edge (Dev) is weak. They have really neutered its ability and length/number of queries you can ask it. It also tries
by roody15 4y ago
Mircosoft's implementation of ChatGPT in Edge (Dev) is weak. They have really neutered its ability and length/number of queries you can ask it. It also tries to sprinkle in some web suggestions as well.
This is worrisome because I think mega corporations like Microsoft and Google want AI to just be a fun little tool for drafting emails and writing jokes and recipes. Any more powerful language constructions are too risky to their existing business models or too risky in terms of PR. People will just screenshot something the AI says philosophically and it will get retweeted and people will lose their mind.
I really hope a open source workable competitor to chatGPT emerges soon.
- toddmorey 4y agoKind of maddening & ironic that we now need an open alternative to OpenAI, considering their original positioning.
- v8xi 4y agoWould love it, too. But at this stage locking it up behind a corporate paywall is probably better IMO - at least until governments around the world are prepared to address a tsunami of misinformation, fraud, and abuse that these systems enable.
- fwn 4y agoThis really does not sound like you would love it to be an open source workable competitor at all. IMO it's not the task of governments to be prepared to address misinformation. It's an individual responsibility. This requires an educated civil society with the media literacy necessary to make a democracy work in the 21st century. It's an ongoing, aspirational project. ...and we can't wait until we think we're ready.
- felixgallo 4y agoThis would be great, if it were only remotely feasible that an individual could ever have the tools and training to counterbalance entire industries of giant disinformation-focused billionaire corporations and governments with thousands of highly trained disinformation professionals using modern tooling, well-funded research departments, and increasingly deep data to spread disinformation. There is no amount of 'media literacy' training (which, in practice, would have to be created and administered by public schools under the watchful eye of some of the selfsame disinformation purveyors, in some states) that will ever be up to the task.
- fwn 4y ago> There is no amount of 'media literacy' training (which, in practice, would have to be created and administered by public schools under the watchful eye of some of the selfsame disinformation purveyors, in some states) that will ever be up to the task. Sometimes I feel that the skill level demanded for media literacy is set artificially high in order to force the conclusion that individuals are categorically incapable of judging their personal information environment. But that does not seem plausible to me. I feel pretty savvy, and I wasn't trained in school. Almost no journalist or public figure has received formalized training in the way you describe. And why would that be necessary, what would they teach beyond common sense? Do you have examples of large-scale disinformation campaigns that aren't sufficiently contextualized through regular mainstream media?
- pydry 4y ago>IMO it's not the task of governments to be prepared to address misinformation. The media wasnt perfect but they were better before the fairness doctrine was repealed.
- freejazz 4y ago> This requires an educated civil society Do you think educated civil societies just spontaneously form, or is it because they are enabled through the government providing things like education (the best way to combat misinformation).
- ImprobableTruth 4y agoOpenAI has been giving this as their reason for years and nothing has happened even though powerful models were released open source. Turns out that misinformation is already incredibly cheap to create and the bottleneck lies elsewhere.
- clarge1120 4y ago> ... at least until governments around the world are prepared to address a tsunami of misinformation The American system, as designed, is diametrically opposed to waiting for governments around the world to protect you from misinformation.
- practice9 4y agohttps://open-assistant.io https://open-assistant.io are working on collecting a dataset and training an open source model comparable to chatGPT
- matheusmoreira 4y agoWe really need the power to run these things on our own computers with 0 restrictions. Not these bastardized corporation friendly versions.
- LadyCailin 4y agoIt requires 300GB of RAM apparently, not sure how you propose to make that work.
- marcyb5st 4y agoAnd specialized hardware. Not sure how long an inference on CPU only will take, but I suspect it will crush usability.
- matheusmoreira 4y agoWell, my laptop has 64 GB. It can't be that difficult to fit 300+ GB of RAM to a personal computer.
- pixl97 4y agoThis is one of those comments you can't tell if its sarcasm or not...
- ojosilva 4y agoWe said the same thing, in one way or another, about indexing and searching the web -- privacy/anonymity, spam/seo, algorithms, bias, corporations.... So today one can build their own crawler and search engine with great opensource software, yet a vast majority is querying a web index through third parties as the computing power required (well, mostly storage and networking) remains brutal, and then all those nitty-picketty details about crawling one needs to painstakingly figure out. Even if by Moore's law we get that great power in our own computers, enough to run a today's chatGPT clone, there will always be a commercial version that is 100x more awesome out there built by actors with immense compute power backed by rivers of corporate money.
- dev-3892 4y agoeverything the ai says is something the ai says philosphically
- yucky 4y agoYou should see the version that the NSA uses :)