[-] kromem@lemmy.world 6 points 4 hours ago

"Shhh honey, I'm about to kill God."

[-] kromem@lemmy.world 2 points 4 hours ago

Unfortunately, removing Harris from the ticket doesn't have the best optics in a lot of scenarios.

[-] kromem@lemmy.world 6 points 4 hours ago

Exactly. The difference between a cached response and a live one even for non-AI queries is an OOM difference.

At this point, a lot of people just care about the 'feel' of anti-AI articles even if the substance is BS though.

And then people just feed whatever gets clicks and shares.

[-] kromem@lemmy.world 1 points 3 days ago

It's right in the research I was mentioning:

https://transformer-circuits.pub/2024/scaling-monosemanticity/index.html

Find the section on the model's representation of self and then the ranked feature activations.

I misremembered the top feature slightly, which was: responding "I'm fine" or gives a positive but insincere response when asked how they are doing.

[-] kromem@lemmy.world 0 points 5 days ago

This comic would slap harder if not for the Supreme Court under christofascist influence from the belief in the divine right of kings having today ruled that Presidents are immune from prosecution for official acts.

That whole divine king thing isn't nearly as dead as the last panel would like to portray it.

[-] kromem@lemmy.world 9 points 5 days ago

But you also don't have Alfred as the one suiting up to fight the Joker either.

[-] kromem@lemmy.world 6 points 6 days ago

This is incorrect as was shown last year with the Skill-Mix research:

Furthermore, simple probability calculations indicate that GPT-4's reasonable performance on k=5 is suggestive of going beyond "stochastic parrot" behavior (Bender et al., 2021), i.e., it combines skills in ways that it had not seen during training.

[-] kromem@lemmy.world 1 points 6 days ago

The problem is that they are prone to making up why they are correct too.

There's various techniques to try and identify and correct hallucinations, but they all increase the cost and none are a silver bullet.

But the rate at which it occurs decreased with the jump in pretrained models, and will likely decrease further with the next jump too.

[-] kromem@lemmy.world 4 points 6 days ago

Here you are: https://www.nature.com/articles/s41562-024-01882-z

The other interesting thing is how they get it to end up correct on the faux pas questions asking for less certainty to get it to go from refusal to near perfect accuracy.

133

I often see a lot of people with outdated understanding of modern LLMs.

This is probably the best interpretability research to date, by the leading interpretability research team.

It's worth a read if you want a peek behind the curtain on modern models.

9
submitted 3 months ago by kromem@lemmy.world to c/technology@lemmy.world
79
submitted 3 months ago by kromem@lemmy.world to c/technology@lemmy.world
[-] kromem@lemmy.world 157 points 5 months ago

More like we know a lot more people that would have zombie bite parties because they "trust their immune system" and simultaneously don't believe in the zombie hoax.

7
submitted 5 months ago* (last edited 5 months ago) by kromem@lemmy.world to c/technology@lemmy.world

I've been saying this for about a year since seeing the Othello GPT research, but it's nice to see more minds changing as the research builds up.

Edit: Because people aren't actually reading and just commenting based on the headline, a relevant part of the article:

New research may have intimations of an answer. A theory developed by Sanjeev Arora of Princeton University and Anirudh Goyal, a research scientist at Google DeepMind, suggests that the largest of today’s LLMs are not stochastic parrots. The authors argue that as these models get bigger and are trained on more data, they improve on individual language-related abilities and also develop new ones by combining skills in a manner that hints at understanding — combinations that were unlikely to exist in the training data.

This theoretical approach, which provides a mathematically provable argument for how and why an LLM can develop so many abilities, has convinced experts like Hinton, and others. And when Arora and his team tested some of its predictions, they found that these models behaved almost exactly as expected. From all accounts, they’ve made a strong case that the largest LLMs are not just parroting what they’ve seen before.

“[They] cannot be just mimicking what has been seen in the training data,” said Sébastien Bubeck, a mathematician and computer scientist at Microsoft Research who was not part of the work. “That’s the basic insight.”

5
submitted 5 months ago by kromem@lemmy.world to c/chatgpt@lemmy.world

I've been saying this for about a year, since seeing the Othello GPT research, but it's great to see more minds changing on the subject.

69

I'd been predicting this would happen a few months ago with friends and old colleagues (you can have a smart AI or a conservative AI but not both), but it's so much funnier than I thought it would be when it finally arrived.

[-] kromem@lemmy.world 196 points 7 months ago* (last edited 7 months ago)

I've seen a number of misinformed comments here complaining about a profit oriented board.

It's worth keeping in mind that this board was the original non-profit board, that none of the members have equity, and literally part of the announcement is the board saying that they want to be more aligned as a company with the original charter of helping bring about AI for everyone.

There may be an argument around Altman's oust being related to his being too closed source and profit oriented, but the idea that the reasoning was the other way around is pretty ludicrous.

Again - this isn't an investor board of people who put money into the company and have equity they are trying to protect.

205
submitted 7 months ago by kromem@lemmy.world to c/world@lemmy.world
12

Pretty cool thinking and promising early results.

11
9
submitted 10 months ago by kromem@lemmy.world to c/chatgpt@lemmy.world

I've suspected for a few years now that optoelectronics is where this is all headed. It's exciting to watch as important foundations are set on that path, and this was one of them.

4
submitted 10 months ago by kromem@lemmy.world to c/machinelearning@lemmy.ml

I've had my eyes on optoelectronics as the future hardware foundation for ML compute (add not just interconnect) for a few years now, and it's exciting to watch the leaps and bounds occurring at such a rapid pace.

[-] kromem@lemmy.world 269 points 10 months ago

The bio of the victim from her store's website:

Lauri Carleton's career in fashion began early in her teens, working in the family business at Fred Segal Feet in Los Angeles while attending Art Center School of Design. From there she ran “the” top fashion shoe floor in the US at Joseph Magnin Century City. Eventually she joined Kenneth Cole almost from its inception and remained there for over fifteen years as an executive, building highly successful businesses, working with factories and design teams in Italy and Spain, and traveling 200 plus days a year.

With a penchant for longevity, she has been married to the same man for 28 years and is the mother of a blended family of nine children, the youngest being identical twin girls. She and her husband have traveled the greater part of the US, Europe and South America. From these travels they have nourished a passion for architecture, design, fine art, food, fashion, and have consequently learned to drink in and appreciate the beauty, style and brilliance of life. Their home of thirty years in Studio City is a reflection of this passion, as well as their getaway- a restored 1920's Fisherman's Cabin in Lake Arrowhead. Coveting the simpler lifestyle with family, friends and animals at the lake is enhanced greatly by their 1946 all mahogany Chris-Craft; the ultimate in cultivating a well appreciated and honed lifestyle.

Mag.Pi for Lauri is all about tackling everyday life with grace and ease and continuing to dream…

What a waste. A tragedy for that whole family for literally nothing. No reason at all other than small minded assholes.

18
submitted 11 months ago by kromem@lemmy.world to c/history@lemmy.world

The Minoan style headbands from Egypt during the 18th dynasty is particularly interesting.

view more: next ›

kromem

joined 1 year ago
MODERATOR OF