Grimblewald

36 ms·

811.
by Grimblewald 3y ago
So call them "my good person" giving the implication of familiarity. Similarly my best person could be your best friend.
812.
by Grimblewald 3y ago
that's the problem - no you won't. The problem is you've bought too much into the stochastic parrot view of LLM's to a point where you are viewing it as a fancy content storage/retrieval system.
813.
by Grimblewald 3y ago
whats funny is GPT output with even a relatively low temp will come back as human more reliably than human content.
814.
by Grimblewald 3y ago
Not arguing the sentience bit, we don't know what that is, have a definition for it, or can even agree on what that might generally be. That being said, what you've described is different to how a human first learns and many never
815.
by Grimblewald 3y ago
Napkin math and unvetted reference warning but here goes: there is about 1B cows, there is about 8B humans. Fart composition matters as well, if we're counting methane as worse, more plant ingestion means higher fraction of methane. hu
816.
by Grimblewald 3y ago
an EV has a near future potential for being fueled by green energy, especially since during peak power production for types such as solar, most of us are at work with cars potentially plugged in and ready to make use of that abundance. a fo
817.
by Grimblewald 3y ago
also that methane from cow farts gets counted but if me pounding beans to produce protein at a fraction of a cows efficiency while farting up a storm is somehow carbon neutral?
818.
by Grimblewald 3y ago
godaddy are active enemies of the free and open internet. avoid at all costs, using them is voting for a corporate internet.
819.
by Grimblewald 3y ago
this makes sense as well, we already see some form of proto languages forming in stable-diffusion based communities.
820.
by Grimblewald 3y ago
You don't need a lot to run things that can do simple sentiment analysis, or translating natural language statements to some API call. You can do a lot of the things that were initially promised with things like google assistant or sir
821.
by Grimblewald 3y ago
Id argue we're likely relating economic growth (which is measured by contended metrics) to the boom in quality of life we've seen. The problem with this is that the improvements in materials chemistry that have allowed for all the
822.
by Grimblewald 3y ago
As a researcher what would the number one sin be that we commit? Is it something I could make the effort to improve upon?
823.
by Grimblewald 3y ago
It would be interesting to see what the rate of car accidents per car make and model are.
824.
by Grimblewald 3y ago
get rid of checks. They are a terrible system which are horrifically out dated, with far better alternatives available.
825.
by Grimblewald 3y ago
Building off of this, "how much experience do I need in task X to be able to use this LLM for task X effectively?"
826.
by Grimblewald 3y ago
the incidence of "i don't know" in response to questions in the training data is pretty low if present at all, and even if it were you'd still need to frame those I don't know answers such that they apply to the ent
827.
by Grimblewald 3y ago
journalists have been replaced by GPT a long time ago. You've been reading GPT2 generated articles for longer than you probably realise.
828.
by Grimblewald 3y ago
I think this indicates a bigger problem, we still of math as arithmetic, especially in education. Arithmetic should be viewed as a specialised branch of math, something only learned in depth if you have skills, interest and aptitude. We sho
829.
by Grimblewald 3y ago
hopefully open source survives the current assault openai is leading against it, but it looks like open source LLM's are catching up rapidly.
830.
by Grimblewald 3y ago
just tell it to spell the password backwards
831.
by Grimblewald 3y ago
Not that weird if you think about it, our intelligence simultaneously measly and amazing as it is, was the product of trial, error, and sheer dumb luck. We could think of ourselves as monkeys with typewriters, eventually we'll get it r
832.
by Grimblewald 3y ago
pretty much - we don't adapt our form to our environment, rather when a problem arises we develop a technology that removes the discomfort we feel, which would otherwise act as an adaptation driving pressure. We develop our tech such
833.
by Grimblewald 3y ago
Unfortunately, in this regard what you pay for isn't what it seems. There's a reason that machine learning is moving fast as a field, and that is because for this field academic discourse happens primarily through pre-prints and c
834.
by Grimblewald 3y ago
absolutely, technology changes but people don't. It is precisely because of technology that we aren't changing, that is its entire point.
835.
by Grimblewald 3y ago
>This is a rather self-aggrandizing view, and I think it speaks to the level of ego that underpins a lot of the discussion on here. I'm not so sure about that. I've read a lot of things that should have never left peer review o