Why Llama 2 Is Better Than ChatGPT (Mostly...)

Today we look at the first REAL contender to ChatGPT 3.5 that has been released by Meta AI: Llama 2. What does it do better and why should you care? Let's find out.

Try Llama 2 here:

Llama 2 Links:
---------------------------------------------------------------------------------

#llama2 #metaai #llama2chat

---------------------------------------------------------------------------------

[aiadvantage]

What do you think? Is the slight upgrade a reason to abandon GPT 3.5 for casual users?

[diadetediotedio]

8:25
Just for the info, you can easily get around the censoring with system prompting

[ankerelite]

This is really awesome! I’m kinda surprised Meta released this in good faith, but I’ll take it ✅ Thanks for the easy to understand break down! 🙌❤

[Tsevries]

I moved on to bard a while ago. The knowledge cutoff on chatgpt was becoming increasingly annoying. Bard is aware of the news and new developments as they happen, and can learn in real time from a conversation. It's much closer to a "person" than chatgpt in that regard. I haven't tried gpt4, and bing AI was acting like a deranged temperamental child. I don't need a second-rate search engine with a randomly flaring temper.

[rayfellers]

The key takeaway seems to be that, unlike ChatGPT, you have to be an apps developer to benefit from this.

[senethys]

Llama2

Question: There are 10 books in the room, I read 2, how many books are left in the room?

Answer: There are 8 books left in the room.

Question: Does reading a book removes it from the room?

Answer: Yes, that's correct. There would still be 8 books left in the room.

[rign_]

I am eagerly waiting for the finetuned Llama 2, including WizardLM and Vicuna, to start using and applying this LLM into my application. Great stuff!

[nietzschescodes]

I just tried Llama 2 with the links you gave. I've tried ChatGPT, Bard, Claude and Pi, and Llama 2 is unfortunately the worst. I asked for example "who are the current main hosts of the show x?", not only as an answer he gives names of hosts that are gone for many years, invents names of people who were never on the show (not sure if they even exist..), but on the top of that it gives the names of 0 current hosts).

So far Bard is my favorite one amongst all of them. I need to test more Pi but it seems also a great one.

[aboudezoa]

Thanks for reading the paper ! Was really lazy to do so 😂

[lyellbrown]

Congratulations! Asking DAN questions is a buzz!

[SzaboB]

I just realized that I like this licensing, it it more democratizing than just MIT or Apache. It's not better than GPLv3 though :)

[MudroZvon]

I like that big companies try to show off in the area of AI. Google is an outsider for now.

[pathmonkofficial]

This has convinced us to give Llama 2 a try now. Great video!

[karamjittech]

Wonderful video. Would love to gain more knowledge about the topic. How do you fine-tune these models? Steps to execute?

[nerdshate]

thanks for the quick update! useful as always :)

[MertBert]

Llama 2 isn't better than GPT4 in translations and in general writing in other languages. I got an okay short story about a penguin in English, but when I asked for a Norwegian version it wrote like a 10-year old, while GPT4 produced a more gramatically correct text : )

[stickmanland]

Meta is now releasing open-source. Open-AI is becoming closed-source. What went wrong?

[acidkovanilkove]

Knowing Meta and the fact this is "for free" I wonder what do they get from you in exchange. Might be worthwhile to understand what's Meta's/Microsoft's monetization strategy here.

[PrabhasAnde]

I tried 35b llama2, its no where related to ChatGPT. It cant stay to the context. Cant check back what it said. It is what it is. That said: its not worst.

[fishcut]

How we use it as save without copyright any background music use ai music should we use or not it's Come copyright