Claude 2.1 is Here - A Real Threat to ChatGPT
Vložit
- čas přidán 13. 09. 2024
- Claude 2.1, the latest version of the AI chatbot developed by Anthropic, marks a significant development in the field of large language models. This new version emerges as a direct competitor to ChatGPT, offering unique features that distinguish it from its predecessors and rivals.
Enhancements in Claude 2.1
The new iteration of Claude includes several notable updates:
Expanded Context Window:
Claude 2.1 offers a context window of 200,000 tokens, equivalent to approximately 150,000 words or 500 pages of text. This allows for the analysis or summarization of extensive documents such as financial statements, code bases, or entire books.
This feature is exclusive to Claude Pro, which requires a subscription.
Reduced Hallucination Rates:
The model significantly reduces the rate of generating incorrect or fabricated information, a common issue in AI chatbots.
Improved Accuracy:
The update has led to a 30% reduction in incorrect answers, indicating enhanced intelligence and reliability.
You can download the list of 100 prompts for Claude here:
skillleap.ai/p...
Master ChatGPT, Midjourney, and top 50 AI tools with Our New AI Education Platform.
Start a free trial Today: bit.ly/skillleap
I love the quality of Claude but there are TOO MANY guardrails. I remember trying to research Nazi and Khmer Rouge history and Claude lost its mind. It would remind me it was an LLM and felt uncomfortable about certain questions. Then, it flat out stopped responding. WTH! It's history!!!! They have an opportunity to takeover during the current OpenAI debacle but they need much more than just increasing the context window to become the dominant LLM.
The only thing that might be more dangerous than true AGI is Woke AGI.
You only get Kosher certified history, goyem. Every single time.
💀
They’re all globalist controlled and thereby blocking any prompts that go against those agendas.
7:30 With decline to answer, they mean that the model doesn't answer to a question they don't know anything about, instead of hallucinating an answer
Testing an AI answer against multiple models is helpful quality control.
The problem with claude is output length which is still very short even after this update.
Thanks for heads up. Seems that censorship is the big problem.
Not censorship, just outright response length. Even for totally benign pdf articles, it responds very curtly when asked to provide a detailed summary.
@@YanivFox exactly. We were using claude api to document our legacy code for the company but the context length is only for input size. The output responses barely reached 4k tokens.
Gpt 4 is just way ahead right now.
@@utkarsh5138In theory, you could bypass the 4,096 token limit for a single output from Claude 2.1 by structuring your queries to receive responses in multiple parts. Each part of the response would adhere to the 4,096 token maximum, and subsequent queries could be designed to continue the conversation or expand on previous answers. This approach requires careful planning of the prompts to ensure continuity and coherence across the multiple responses.
Hi there!
My question that is it possible to generate a 5-10 A4 pages summaries from a long document?
Can you create your own personalized chatbot with Claude? I'd like to feed it documents which it stores, to be later accessed by team member if requested
You ask a weird question. Not sure how feeding it documents is related to making a personalized chatbot. Claude 2.1 is the best program for personalized chatbots. Insanely better than GPT 4. As for documents, I wouldn't know, but the video explains. It depends on how big they are. I've used NotebookLM for the documents.
@@HiddenPalm Maybe you just dont get the question, seems like your problem
I had a paid plan, but the actual upload data size limit was no where near what was advertised. Hope that has changed.
2:00 GPT4-turbo has 128k token context window
Yea that’s true. I meant to say gpt 3.5
I will still use Undetectable AI as my paraphraser tool, because it humanize my essays to avoid detectors
20$ is expensive for Claude and limited massage ?!!
it should be around $10 in my opinion, however I'm on pro. Still thinks what it outputs, the amount they are charging is bit pricey 😌
Interestingly, check the end of the video again at the order of the 10 listings by market cap… Some are out of order.
Did you check 2.1 for the hallucinatory effect about that company that you found with 2.0
I just tried it. It said it doesn’t have any specific information on that company.
Tx. So as predicted
What use cases do you use for Claude 2.1, GPT4, and Bard? Do you use each of these models for different use cases?
I like Claude for data analysis, Bard for up to date research that requires citation and GPT for everything else.
I have several of these AIs hooked up in my browser. I’ve not used Claude that much yet. I mostly use the web-enabled AIs., I’d say mostly Bing, Phind and Perplexity. I find Phind best overall for programming and tech tasks.
gpt 4 vision (I think api only) is currently 128K tokens
asked Claude to summarize a CZcams video. Did not do it.
The link is not working?
Should be fixed. if not, email me at support@skillleap.ai and I can email you the PDF
Bro, your light backgrounds are torture for us with large screens. Ouch ...
I don't see a dark more on Claude. My ChatGPT and Bard and my OS are all on dark mode.
in need the pro version but in Switzerland is it not possible to buy it :-(
VPN
@@Eu.Ochiuzzo i know but switzerland is not on the location list
@@fabriziocasula perplexity.ai
Thank you for your best advice
Proble mwith claude is that its way too restrictive.
Is there any 3rd party tool (ex. Raycast) for Windows programmers? I'd love to feed Calude 2.1 or ChatGPT4 a set of data, and say plot the csv using matplotlib where ...
The only services I know that'll give monthly unlimited llm hit is pro plan perplexity. But the model you could use is only llama2 70b.
It's noticeably faster on my end. Hopefully it won’t make stuff up as much when I ask it to print out something we just cowrote word-for word.
😢
Nice!
Well... It's broken now. Don't know for how long.
Have you noticed that Claude now produces very poor content. It's basically unreadable - very convoluted, like a meaningless machine text. It happened after New Year. What could be the problem?
It's broken dude. Won't write anymore.
I think Bard is better than the two
No
lol
Ha I luv how you said you get all kinds of errors when you upload a word .doc! Of course because it’s from Microsoft where nothing positive has ever come from them
You cannot use it in EE, that's a joke! Such an improvement and such a failure. Joke!
Love your content from Fatick Sénégal West Africa 🎉 First
nice!!!
First viewer
I uploaded an imperfect .jpg of a page from a German tax document. It did the translation fine. You may want to experiment with that sort of thing.
Who is better for creative writing, claude 2.1 o chatgpt plus? @appoftheday #appoftheday