I was going to ignore this as a troll comment because computer memory has its antecedents in human memory but the commenter is right - the combination of memory and cache to talk about human memory seems misleading.
893 karma · joined August 21, 2020
> This website is using a security service to protect itself from online attacks. The action you just performed triggered the security solution. There are several actions that could trigger this block including submitting a certain word or phrase, a SQL command or malformed data.
Could a tighter operating range specified in the system prompt along with lower temperature bands which cause less output variability help?
Side note - I saw upthread that you were looking to rebrand “prompt injection”. I propose “behaviour induction” or “induced behaviour”
Still, there’s no city like Mumbai in the whole of India. It’s literally got something for everyone.
Not a CPA so I’m pretty sure I’m oversimplifying this and there are lower bounds I don’t know about which prevent most gig workers from operating like this.
Making data ingestible is just laborious and unappetising to an engineer. So hand it over to a data entry operator who’ll do this happily for a wage.
I’m not sure why the issue of making data available to an LLM is a failure of the LLM.
Btw, OpenAI’s LLM can very well parse images, read handwriting, fix sloppy spellings, and extract objects from unstructured data.
In any case, it doesn’t matter a whit if the average is one meltdown if your kid has gone without two consecutive meals. All this theorising goes out the window when the third meal is coming up and you’re struggling to get a morsel inside your kid. How many consecutive meals are you willing to let your kid skip on principle?
Comparing what we did in the past, how we lived in the past, what worked two decades ago is basically useless.
Phones exist. On demand entertainment exists. Kids know it. You know it. You aren’t putting any genie back in the bottle by trying to explain to your kid that since you grew up without being amused into eating, they need to do that too.
Being so black and white about sticking a video in front of a kid to make sure it sits in one place and eats is dismissive of parents who have to make these decisions in real time.
Do we know those kids don’t get play time with friends? Or down time to get bored and hence be forced to invent games in their own heads? We only see and opine on “bad” behaviour, not the what happens behind closed doors the remaining 23 hrs of the day when the kid is out of our line of sight.
C’mon Kyle, shows us what’s behind the choli, dude!
There’s even a name for this bias - Baader Meinhoff effect
Since GPTs are not deterministic, any intelligence we attribute to it relies on the observer/attributor.
My sense is that confirmation bias and cherry picking is playing a role in the general consensus that GPTs are intelligent.
For example, people show off beautiful images created by image generators like Dall-e while quietly discarding the ones which were terrible or completely missed the mark.
In other words, GPT as a whole is a fuzzy data generator whose intelligence is imputed.
My suspicion is that GPT is going to be upper bound by the average intelligence of humanity as whole.
His second wife apparently asked him to buy Twitter and fix its, in her opinion, liberal bias.
An employment is a contract which both parties enter into willingly. Termination of contract deserves some level of empathetic glad handling, however minimal. It's just game theory - if you plan to hire again, you have to be gracious while firing someone because word gets around.
Don’t think Sam or Greg have it in them to build a competing AI model suite, that too inside a bureaucracy like Microsoft.
I think this is exactly what OpenAI wanted - get the business types out and focus on building brilliant models which asymptotically approach AGI whose safety and ethicality they can guarantee.
For example, i have an assistant which is supposed to parse an uploaded file and extract useful info from it. To use this assistant, I create a thread and a run and attach it to the assistant with a different file-id. About half the time, the assistant simply throws up its hands and says it can’t parse the file I supplied with the thread. Retrying a few times seems to do the trick.
1. Sam gets the company to agree to pick up the costs of lawsuits relating to unauthorized not-fair-use use of copyrighted content. 2. Announces this. 3. Then tells the board that he is directly on the hook on some lawsuits - essentially creating a safety net/insurance policy for himself without being candid with the board.
Hence, fired.