AI startup Anthropic, backed by Google and hundreds of millions of venture capital, today released the latest version of its GAI technology (generative artificial intelligence) Claude. The company claims that the AI chatbot can surpass OpenAI's GPT-4 in terms of performance.
Anthropic's Claude 3 model series: Claude 3 Haiku, Claude 3 Sonnet, and Claude 3 Opus, with Opus being the most powerful. Anthropic said that it is compatible with ChatGPT and GPT-4 as well as Google's Gemini 10 ultra and other models, with enhanced analytics and performance on specific benchmarks.
Claude 3 is Anthropic's first multimodal GAI that analyzes text and images, similar to GPT-4 and Gemini. Claude 3 can also handle drawings, charts, graphs and technical charts, PDFs, slides, and other document types.
One of the more distinctive features of Claude 3 is its ability to analyze multiple images (up to 20) in a single request and compare them at the same time.
But there are limitations to Claude 3's image processing capabilities, and the company admits that Claude 3 is prone to errors when processing low-quality images (less than 200 pixels) and struggles with tasks involving spatial reasoning (such as reading analog clock faces) and object counting (Claude 3 cannot give accurate results for the number of objects in the image).
AI drawing.
Claude 3 also does not generate artwork, these models are rigorous image analysis.
Anthropic said that Claude 3 is better able to follow multi-step instructions than other versions of GAI, generating structured output in formats such as JSON, allowing communication in non-English languages, whether working with text or images. Anthropic said that Claude 3 should also reduce the frequency of refusal to answer questions due to a more nuanced understanding of instructions. The model will reference the answer to the question, increasing the answer.
Compared to Anthropic's older models, the Claude 3 is easier to cue and boot. Achieve the desired results with shorter, more concise prompts. Some of these improvements stem from the extended context.
The context or context window of a model refers to the input data (such as text) that the model considers before generating the output. Models with smaller contextual windows tend to "forget" recent conversations, causing them to go off-topic – often in questionable ways. Large context models can better grasp the narrative flow of the data they receive.
Anthropic said that Claude 3 will initially support a context window of 200,000 tokens, which equates to about 150,000 words, and some customers will get about 700,000 words. This is in line with Google's latest GAI model, Gemini 15 pro quite.
Now, just because Claude 3 is an upgrade from the previous version doesn't mean it's perfect.
Anthropic admits that Claude 3 is not immune to the problems that plague other GAI models, namely making things up. Unlike some GAI models, Claude 3 is time-sensitive and can only answer questions using data prior to August 2023. Although Claude speaks several languages, it is not fluent in some minor languages compared to English.
Anthropic believes that the model's intelligence has not yet reached its limits, and the company plans to enhance the features of the updated Claude 3 in the coming months.