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

Phind Model beats GPT-4 at coding, with GPT-3.5 speed and 16k context

phind.com

Post reply on HN
Print view View original

351–358 of 358 posts

Re: Phind Model beats GPT-4 at coding, with GPT-3.5 speed and 16k context

#351

Post by slowhadoken »

make3 wrote: post #340
slowhadoken wrote: post #256

Earlier quoted context omitted.

I've heard this coldtake before but OpenAI's source code isn't open to academic scrutiny. So I don't understand why some people are so confident about how it works. It's certainly not magic and Phind seems to be capable of it citation.

It's transformer based language-modelling 101, not really a take, just stating facts. It's highly unlikely Phind has completely fundamentally changed all the exact same problems that the whole field is working on simultaneously, single-handedly, in a purely novel way. It's just how transformers work.

Phind appears to be doing it though. LLMs are stochastic parrots, I don’t see a radical difference. Input goes in, output comes out. Neural network aren’t magic, they’re a complex function. 1 node or a billion we can track the data that’s changing the weights inside the network.

Re: Phind Model beats GPT-4 at coding, with GPT-3.5 speed and 16k context

#352

Post by dontreact »

letitgo12345 wrote: post #279
dontreact wrote: post #192

Earlier quoted context omitted.

def calculate_relative_advantage(interest_rate, loan_size, market_return): # Your calculation logic here pass Chat gpt actually implements it

Just prompt it to implement the function

I did and it was wrong. I was responding to the claim that they got a plausible result.

Re: Phind Model beats GPT-4 at coding, with GPT-3.5 speed and 16k context

#353

Post by make3 »

slowhadoken wrote: post #351
make3 wrote: post #340

Earlier quoted context omitted.

It's transformer based language-modelling 101, not really a take, just stating facts. It's highly unlikely Phind has completely fundamentally changed all the exact same problems that the whole field is working on simultaneously, single-handedly, in a purely novel way. It's just how transformers work.

Phind appears to be doing it though. LLMs are stochastic parrots, I don’t see a radical difference. Input goes in, output comes out. Neural network aren’t magic, they’re a complex function. 1 node or a billion we can track the data that’s changing the weights inside the network.

I do this for a living

Re: Phind Model beats GPT-4 at coding, with GPT-3.5 speed and 16k context

#354

Post by ewokone »

xrd wrote: post #57

I know it isn't popular, but I wish there was a way to use this inside Emacs. Or, vim. I just don't want to use VS Code anymore.

I have been a vs code power user and switched to pycharm two years ago and will never go back because of the features for working with multiple environments and projects in pycharm.

Working with phind needs to be available in pycharm for me considering switching from gpt4 to phind. Chatting with phind on my local files is the feature I am looking for.

Re: Phind Model beats GPT-4 at coding, with GPT-3.5 speed and 16k context

#355

Post by sounds »

selfhoster11 wrote: post #306
sounds wrote: post #183

Earlier quoted context omitted.

Can we have a larger discussion about the tradeoffs that come with open sourcing a model? When fb released Llama they obviously gained a huge amount of developer goodwill but it also required them to invest a serious amount of their own developer time to engage with the community. I'm asking the community what it can offer the company? Or is this just self-abnegation by the company that releases the model?

I question the word "required". They, or anyone else releasing an open source product into the world, doesn't owe anyone anything, least of all support. As long as there are enough instructions to run the thing, you are perfectly within your rights to let the community sort out the rest between themselves.

I've noticed that even though "They don't owe anyone anything," the community doesn't actually adhere to it. If they shove code over the wall like FAANG companies do now, it appears to upset the community, who will then treat them with hostility.

Re: Phind Model beats GPT-4 at coding, with GPT-3.5 speed and 16k context

#356

Post by gnaritas99 »

slowhadoken wrote: post #351
make3 wrote: post #340

Earlier quoted context omitted.

It's transformer based language-modelling 101, not really a take, just stating facts. It's highly unlikely Phind has completely fundamentally changed all the exact same problems that the whole field is working on simultaneously, single-handedly, in a purely novel way. It's just how transformers work.

Phind appears to be doing it though. LLMs are stochastic parrots, I don’t see a radical difference. Input goes in, output comes out. Neural network aren’t magic, they’re a complex function. 1 node or a billion we can track the data that’s changing the weights inside the network.

[dead]

Re: Phind Model beats GPT-4 at coding, with GPT-3.5 speed and 16k context

#357

Post by NoNoisle »

alex-moon wrote: post #343
geedzmo wrote: post #308

Earlier quoted context omitted.

I don't see how this would be relevant for a code model? The code model isn't trained to retrieve papers/articles, it's meant to complete code. Whether or not you find hallucination in a unrelated task isn't particularly interesting.

Damn, this is how I learn that HN doesn't have a block function. What a shame. My friend, can you do me a favour and actually click the link and have a play with the app? If you do, you will discover that what you're dealing with there is an LLM. That's literally why it's being compared to other LLMs. No idea what you were trying to achieve with this comment. "The code model isn't trained to retrieve articles." a) ne…

Lmao

Re: Phind Model beats GPT-4 at coding, with GPT-3.5 speed and 16k context

#358

Post by razodactyl »

rushingcreek wrote: post #220
drcode wrote: post #219

Earlier quoted context omitted.

that would put them significantly ahead again, for my use cases

We will eventually increase the Phind Model to 100K tokens -- the RoPE embeddings in Code Llama were designed for this.

What about ALiBi and Sliding Window Attention?

Additionally Apple researchers seem to be playing with "Attention Free" variants.

Post reply on HN