Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

MiniGPT-4

minigpt-4.github.io

Post reply on HN
Print view View original

151–160 of 337 posts

Re: MiniGPT-4

#151

Post by f1codz »

ikurei wrote: post #109
sho wrote: post #14

Earlier quoted context omitted.

> they're doing something really simple -- take BLIP2's ViT-L+Q-former, connect it to Vicuna-13B with a linear layer, and train just the tiny layer on some datasets of image-text pairs Oh yes. Simple! Jesus, this ML stuff makes a humble web dev like myself feel like a dog trying to read Tolstoy.

> This ML stuff makes a humble web dev like myself feel like a dog trying to read Tolstoy. Just like any discussion between advanced web devs would make any humble woodworker feel? And just like any discussion between advanced woodworkers would make a humble web dev feel? "It's really simple, they're just using a No. 7 jointer plane with a high-angle frog and a PM-V11 blade to flatten those curly birch boards, then a…

You make a good point. Except that a number of these concepts and tooling in the ML world have been slingshotted into the forefront in a relatively short time and it has been hard to play catch up. For eg. - someone said "frozen Vicuna" below - what does that mean?

Re: MiniGPT-4

#153

Post by collaborative »

Uehreka wrote: post #61
collaborative wrote: post #25

Earlier quoted context omitted.

Adding GPT4 to anything now increases marketing x4. So many AI news coming out lately that not adding it risks drowning in a sea of info.. even in the case of a good project

The word for this is “trademark infringement”. You are specifically not allowed to capitalize on the marketing of another entity’s product to bolster yours by implying through your name that you are somehow related. This is why “DALL-E Mini” had to change their name to craiyon.

At this point the letters GPT make more sense than "AI" or "LLM" in many peoples minds

Re: MiniGPT-4

#154

Post by lxe »

Why Vicuna over something like oasst-pythia? Would love to see a table comparing all the new models side by side.

Re: MiniGPT-4

#155

Post by brycedriesenga »

dceddia wrote: post #111
moffkalast wrote: post #69

Earlier quoted context omitted.

DriveGPT as it hits 2 parked cars and runs over a dog: "Apologies for the confusion earlier. You are correct that the cars were in the planned local planner path..."

Sorry, yes, that was a dog. Let me try again and this time I will definitely not hit anything. Sorry, that was another dog.

"Was that another dog?"

BingDrive: I'm sorry, but I prefer not to continue this conversation.

Re: MiniGPT-4

#156

Post by dalys »

dom96 wrote: post #62

Can this decode CAPTCHAs? Are we finally in a place where CAPTCHAs are useless?

I was also thinking about this the other day.

Idea one: Captchas are to become pretty useless as a "is this a human" tactic soon. Maybe it already is, I don't know. What other things could we think off to prove someone is human? I was watching Lex Fridman and Max Tegmark and they were remarking on how Twitter using payment as a differentiator between human and bot is actually really good. And maybe the only way we can reliably determine if someone is a human or not right now. Just by the virtue that having thousands of bots doing something, that suddenly costs $5 per event will deter most attacks. Integrating online identification systems from various countries could be one tactic (such as https://en.wikipedia.org/wiki/BankID that we use in Sweden to log in to basically any online service). New startup: Un-botable authentication as a service.

Idea two: Since captchas are useless, we'll be able to do bots that can do almost everything on the web. No need for writing automation scripts, headless browsers, regexp etc. Just feed real visual data from browser to GPT-4 (or MiniGPT-4 or similar). Give instructions like "You need to accomplish this task: Go to facebook.com and create a user account and be friends with 100 people and act like a human. Follow the instructions on the website.". Then let the bot figure out where to move the mouse and send click events, keyboard events etc. Obviously much slower than a scripted bot, but would be very hard to detect as a bot. Good for E2E testing also? New startup: WebGPT: Bots that can do any task on any website. TestGPT: E2E testing service. Just write test instructions like "Make sure the login works on /login" and you're done! And you thought flaky tests were a thing of the past... Would be kind of cool for UX testing also. "You are a 43 year old truck driver in Spain, please go to news.ycombinator.com and see how easy it is for you to register an account, log in, go to any news story and make a new comment. Report anything you think is confusing."

Re: MiniGPT-4

#157

Post by macrolime »

thewataccount wrote: post #127
macrolime wrote: post #90

Earlier quoted context omitted.

The A6000 is actually the old generation, Ampere. The new Ada generation one is called 6000. Seems many places still sell A6000 (Ampere) for the same price as RTX 6000 (Ada) though, even though the new one is twice as fast. Seems you can get used RTX A6000s for around $3000 on ebay.

That.... That explains why I can't find it and makes a ton of sense..... I think that's such a silly name for it, but oh well Thanks for the correction!

Just to add to the confusion, there's another older RTX 6000 with 24GB ram. This is from an even older generation, same as the GeForce 20 series.

Re: MiniGPT-4

#158

Post by ttul »

ikurei wrote: post #109
sho wrote: post #14

Earlier quoted context omitted.

> they're doing something really simple -- take BLIP2's ViT-L+Q-former, connect it to Vicuna-13B with a linear layer, and train just the tiny layer on some datasets of image-text pairs Oh yes. Simple! Jesus, this ML stuff makes a humble web dev like myself feel like a dog trying to read Tolstoy.

> This ML stuff makes a humble web dev like myself feel like a dog trying to read Tolstoy. Just like any discussion between advanced web devs would make any humble woodworker feel? And just like any discussion between advanced woodworkers would make a humble web dev feel? "It's really simple, they're just using a No. 7 jointer plane with a high-angle frog and a PM-V11 blade to flatten those curly birch boards, then a…

The thing is, machine learning sorta requires a few math prerequisites: linear algebra, differential equations, and to some degree vector calculus. Most web developers don’t have this background.

Re: MiniGPT-4

#159

Post by ikurei »

wnolens wrote: post #143
SpaceL10n wrote: post #130

Earlier quoted context omitted.

Web devs have become blue collar!? =P Great idea, actually. I do hope for a curriculum that enables kids on the trade school path to learn more about programming. Why not Master/Journeyman/Apprentice style learning for web dev??

That's kind of how I think about bootcamps pumping out web devs. They're like trade schools, teaching you just enough fundamentals to know how to use existing tools.

I kind of agree, but I'd add that I don't think it's a bad thing.

Re: MiniGPT-4

#160

Post by GaggiX »

lxe wrote: post #154

Why Vicuna over something like oasst-pythia? Would love to see a table comparing all the new models side by side.

All Pythia models were trained on 300B tokens, LLaMa models were trained on 1/1.4T tokens.
Post reply on HN