Claude 3 ai just released.

byTheCatherine

14K followers

2 months ago

CLAUDE 3 was released yesterday by Anthropic. I had to test it as I have a paid account and it suddenly appeared.

Today I started exploring the new features listed below

1. A new standard for intelligence: Claude 3 Opus, the most advanced model, outperforms its peers on various cognitive tasks, exhibiting near-human levels of comprehension and fluency [1].
2. Near-instant results: The Claude 3 models can power live customer chats, auto-completions, and data extraction tasks with immediate and real-time responses. Haiku, the fastest model, can read information-dense research papers in less than three seconds [1].
3. Strong vision capabilities: The Claude 3 models have sophisticated vision capabilities, allowing them to process a wide range of visual formats, including photos, charts, graphs, and technical diagrams [1].
4. Fewer refusals: The Claude 3 models show a more nuanced understanding of requests and are significantly less likely to refuse to answer prompts that border on the system's guardrails compared to previous models [1].
5. Improved accuracy: Claude 3 Opus demonstrates a twofold improvement in accuracy on challenging open-ended questions compared to previous models. It also provides admissions of uncertainty instead of providing incorrect information [1].

The first thing I did was ask it for an outline for an ebook for women who were preparing to be a digital nomadess. I was impressed with the outline and the speed. . But then I tried Partha's trick to ask it to improve and expand.

I was blown away, normally I ask four AI platforms for an outline of anything an article or an ebook and then combine them. I normally use Genesis, Chatgpt 4, Claude and WA AI. I honestly didn't think I had to with this outline.

I will need to try it a bit more to be sure if point 1 is accurate about its levels of comprehension , but the firt results are encouraging and it is lightening fast.

next I tested what it could do with uploaded images. I uploaded a jpeg of hash browns that I asked Chatgpt 4 to create. I entitled the image Guatemalan Hash browns with a salsa picante. A salsa picante is a red spicy sauce. I asked Claude to create a recipe based on that photo. I asked it to use Guatemalan spices.

It surmised that it was hash browns with colelsaw and gave me a recipe for both. The hash browns had too much egg in it. I am an ex chef so I picked up on that immediately. Of course I could have asked it independently for a salsa rojo or a salsa verde recipe, but I didnt. The test was how well it interpreted the image. On balance it was fifty percent successful but I certainly wouldn't have wanted to eat the results.

Point 4 fewer refusals Well it failed on that one. I asked it for guidance on how to become a digital nomadess and it gave me a three paragraphs refusal that boiled down to.

After my upgrade iti s unethical for me to explore becoming a digital nomadess. To be fair I just stared at it in amazement. Today it did deliver an excellent job, I can be sure it wasn't exactly the same prompt as I wrote both from scratch . I can be sure both prompts were thorough in explaining what I wanted.

So the jury is still out on point 4. As to point 5 it wasn't certain in its refusal yesterday. Obviously my next prompt was "why do you surmise that being a digital nomadess is unethical " its answer was it was not programed to copy copyrighted material.

Overal it is fast, that is for sure and it so far appears to be more accurate and comprehensive than Claude 2. However, one thing it still hasn't got is a web search. So I will still pay for three different AI models. For now.