I've seen LLMs do something correct 98% of the time then randomly do something crazy that a human would never do because we have continual learning
As humans we don't have our memory reset multiple times per day
As humans we don't have our memory reset multiple times per day
I've seen humans vote for Brexit, re-elect Trump, ask questions clearly already answered in an FAQ, try to pull on a door labelled "push", and insist on giving me homeopathic silicon dioxide pills* that cost £5** for a 10-12 gram packet.
Continual learning is a difference, but not by itself a reason to care about "deterministic results".
Nor, indeed, correct results.
> As humans we don't have our memory reset multiple times per day
Humans need sleep well before they can read a million tokens' worth of written text. We're more like 300k tokens if you're actually reading and not skimming for 16 hours straight.
Again, different (in soooo many ways), but this isn't a relevant difference when the topic is "deterministic results".
* yes, sand: https://dailymed.nlm.nih.gov/dailymed/fda/fdaDrugXsl.cfm?set...
** and that was what it cost in the 90s
There is no RNG involved when I decide to push vs pull the unlabeled door to my building every morning, it becomes deterministic because its baked into memory
You can put stuff in context to deal with this but you can't do that for everything, its not practical and you would blow the context window
You, me, everyone, we all fail tasks with some non-zero probability, including tasks we've done many times before. This is literally why typos happen. We don't always catch the typos we make, this is literally why spellcheck exists. We don't always catch logical errors etc., this is literally why professional writers work with professional editors. We don't always spot errors in code we write; this was why compiler errors have to exist at one level, why unit tests and integration tests exist at another, and why despite automated tests we have QA departments.
Statistically speaking, you as a human will have made more comprehension errors reading that paragraph than even LLMs a few years back.
Not zero, just fewer. They're not magic and neither are we.
> There is no RNG involved when I decide to push vs pull the unlabeled door to my building every morning, it becomes deterministic because its baked into memory
There's a lot of random inputs into human behaviour. How much they become random outputs is a similar question as for AI. But like I said, typos are a trivial existence proof of humans making randomised errors on a thing they've done many times before. Timing error at random and a "the" becomes a "teh".
Here's one for gross failure modes being random: Small (and apparently random) groups of neurons in your (and all animal) brains briefly enter sleep-like states even while you remain apparently awake. This starts well before you even feel tired. And like I said, looks to be random.
Your memories are not "baked" at any point; they're made of and by cells doing chemistry. Your entire body is a trillion small bags of chemicals that communicate with each other by leaking out signalling molecules and electrochemical gradients into each other's water supply, and this goes wrong sometimes. Human memory is known to be fallible, and also manipulable after the fact.
In the sense that if I hire a competent person, I can tell them something and they will deterministically do it
For an LLM I have no mechanism to even update the weights
All you can do is play around with context, which is like taping a post-it note to someone's desk
Imagine you had an employee where you had to tape 1000 post-it notes to their desk