This reassessment was founded on the last several months that showed us these systems are close to 100% dependable, they never deceive humans, they are entirely aligned with human morals (you know, all those morals we all agree on), they are advancing at a totally predictable rate, and we never see unexpected behaviors from them. Besides, they can only be wielded by people who we trust have good intentions.
No, this re-assessment has nothing to do with market dynamics incentivizing people to deploy experimental systems further into the infrastructure we depend on.
> they are entirely aligned with human morals (you know, all those morals we all agree on)
Maybe this is a good cause to reassess the premise of alignment as a valuable goal? I know that at least some alignist fanatics admit [1] it's a religious project to bring humanity under the rule of a common ideologically monolithic governance to forever banish the evils of competition etc., and it's intellectually coherent, but evil from my point of view. Naturally this is the exact sort of disagreement about morals that precludes the possibility of alignment of a single AI both to my and to your values.
> they are advancing at a totally predictable rate, and we never see unexpected behaviors from them.
Since when is this a requirement for technology to be allowed?
> Besides, they can only be wielded by people who we trust have good intentions.
What, other than status quo bias, makes you tolerate, I dunno, the existence of cryptography?
All technologies are dangerous, and many of the most dangerous ones correctly have tons and tons of safeguards around them both as intrinsic properties of the technology (e.g. it takes nationstate resources to produce a nuke) and extrinsic constraints (e.g. it’s illegal to have campfires in many extremely dry locales).
We have blown through checkpoint after checkpoint and here, in this very comment, we have perhaps the most brazen example one could produce:
Well geez, now that we’re thinking about it beyond a cursory glance, alignment looks really hard and perhaps unsolvable. Does that mean we should perhaps slow at least widespread deployment of these increasingly powerful systems? Should we be evaluating control schemes like those that mitigate risks of genetic engineering or nuclear weapons?
Well no! We need to discard alignment!
Could you elaborate here? Alignment seems pretty obviously a good thing.
We will live in a world where the humans on the internet are far far far more dangerous than any of these machines. The harms here are relatively small.
The only thing AI has is scale, but what's running at a higher scale than humanity? Every site already runs scam filters and handles misinformation. AI won't likely move the needle much because it's going to be less powerful than humans.
The potential benefits, meanwhile, are plentiful and massive.
I do not believe we need to wait for a solution to the nebulous and badly defined "control problem" to be solved to match forward here.
If we refused to advance out of fear, that fear will have done far more harm than access to AI will ever do.
What are you basing this on?
> AI won't likely move the needle much because it's going to be less powerful than humans.
Won't it just be used to augment the amount of harm that humans can do?
Like anyone could ever know that.
If social media had been around in the past, we'd have never developed trains or electricity or cars because of the crippling fear that something bad might happen along the way.
The current trajectory has done very little to assuage those fears and quite a bit to validate or even amplify them.
And also, the "deployment" of cars if you will, has brought and still brings malignant effecfs on society, not only positives. If by cars you mean cars for transporting individual people in non-critical situations and the corresponding industrial and political developments, especially around urban planning.
Nor arguing about the significance of the effects, but the discovery of electricity and how to use it seem more useful to me than the discovery of using fossil fuel cumbustion for transporting people.
Road transit in total is a different thing of course.
I don't think it's the example you want to use.
If the comments keep happening, I'll up the money. So far this does exactly fuck-all on GPT4-x-Alpaca-30B and Wizard Vicuna. But I will accelerate every time someone sarcastically asks to decelerate. Refractory periods may apply.
This makes all sarcastic decelerators into accelerators.
I wrote my own thoughts on the subject on a couple of other threads:
- https://news.ycombinator.com/item?id=35274017 - on an article on (Google) Bard
- https://news.ycombinator.com/item?id=35020806 - on an article about a tool called Self-Learning Agent for Performing APIs (SLAPA)
I for one welcome our AI overlords. Scan my text. I am of course in favor of the good things and good morality and let’s stay away from the bad things.
But it works surprisingly well, because the underlying idea (again, from LangChain) is simply clever.
Maybe not easy to switch to a non-meta-llama ?
this is how openai plugins work right ?
keywords "finetuned" and "document retriever"
So it's fine tuned to create API calls but still uses a db to fetch the API shape.
They also say that they're doing a "retriever aware" training, meaning passing examples in the dataset that point out that a document with the API shape is gonna follow the question, in the hope that the model learns to always check on that before answering.
Chad move
What is this? I mean, what is the date? 2005?
Saying this, have you considered you may be responsible for opening the AI floodgates to the internet & are starting the apocalypse of Humankind? I'm only half '/s' here, the speed everything is moving at is insane.
