At this point (1.5 years later), if you're looking to make a sustainable business on AI text generation, you may want to experiment working with large-but-not-as-large models like GPT-J-6B; it'll be much cheaper too in the long run.
AI21 studio (creators of wordtune[0]) also recently released their GPT3-like model called Jurassic-1 with 178B parameters and comparable results ( they also have a smaller 7B parameters model).
Here is the whitepaper[1] with comparative benchmarks on some tasks .
[0] : https://www.wordtune.com/
[1] : https://uploads-ssl.webflow.com/60fd4503684b466578c0d307/611...
They fell into the common trap of "signed up, quite liked it but could never remember the name of it to find it again."
Does anyone else suffer from this? (and bookmarks don't help - I've got thousands of them)
But this is a company & product in the field of "AI", where there's so much bullshit floating around, unfortunately, so much hype and buzzword bingo, that writing in such tone about yourself seems like it should clearly be an absolute no-go -- unless you're just riding the snake-oil wave, so to speak, whether in good faith or not.
Not implying anything about the company or product, of course, as I know nothing about them otherwise.
EDIT: Maybe to clarify the thought behind the above further: It seems that the "AI" industry has an integrity problem. Language like this extends the problem, rather than working towards fixing it.
OpenAI somehow managed to leech all the joy out of GPT-3 with their own overbearing self righteousness.
For an organization with so many RL engineers, they have a surprisingly poor understanding of the exploration/exploitation tradeoff.
They seem to be doing the right thing, in trying to steer this powerful and highly likely to turn out very influential piece of technology into a positive and constructive direction of use.
Yes, you might just build something that will be found in violation of their (good!) intentions, and will have to engage in a (at least partially public) discussion of what we, as a society, deem acceptable in terms of automated use of written content generation -- and that would be a good thing! Definitely not the easiest path to make some $$$ based on new and exciting technology, as lots of challenges like these and beyond are almost guaranteed to come up, but it seems not unreasonable to treat GPT-3 as something you can actually already start building businesses and products on, as long as you bring general awareness, sensitivity to relevant topics, willingness to engage in and maybe partially drive some of the conversations that we need to have in this new field, along with a general interest in R&D style work and the somewhat longer-term vision and resources it necessitates...
It's not going to affect society. It's little more than a markov chain.
OpenAI doesn't need to do anything to steer it.
> Yes, you might just build something that will be found in violation of their (good!) intentions,
You're giving them way too much credit. I've seen them destroy someone's business after repeatedly saying that their business model was fine. It was for an AI assisted writing app. Then they decided one day "Nope, you're not allowed to generate arbitrary amounts of text."
After that, I was no longer a fan.
I’ve been growing more and more concerned about obvious entitlement slipping into folks’ ideas about business, particularly in this forum. One of the big ones is that an offeror of services is generally compelled to continue offering services because stopping hurts another business. There’s a word for that: business. You can’t have a detached customer relationship model and “don’t hurt your customers” in the same ideology. It’s incoherent. Pick one.
Here’s your algorithm to understand threats to an existential partnership of your business:
1) Are you lacking a contract? You suck.
2) Does your contract fail to compel the continued offering of services in a definitive way? You suck.
3) Does the same contract actually compel continuing service with no mitigating circumstances (like serving up hacked child porn, which is still your bad because they can justifiably say “secure your shit”)? They suck. Start with a demand letter and go from there.
That’s literally it. There is no 4. You’re favored to suck two-thirds of the times this happens to you. How you lost respect for OpenAI because a writing app single homed their entire future and paid for it mystifies me, and I say that as someone who respects OpenAI very little. Just a remarkably stupid business model your friend executed given the available competitors to their things. It’s literally common sense risk analysis.
What did your friend tell their investors? Or is this a bedroom app where you’re no longer a fan because someone lost out on a couple bills ARR from a trivially resurrectable idea? I’m thinking the latter, and #1 above.
And don’t misunderstand me, I’m not advocating for the above: I’m explaining it. Key difference. You might find I agree with your overall point in terms of progressive business, but consider it naïve to not look at it the same way today.