Anthropic, an AI company backed by Google, recently introduced the latest version of its chatbot technology, Claude 3 (TechCrunch). Anthropic claims Claude 3 outperforms OpenAI’s GPT-4 in various aspects.
Claude 3’s Strengths:
Family of models: Claude 3 comes in three tiers (Haiku, Sonnet, Opus) catering to different needs and budgets, beginning with a free (but somewhat limited) starting option.
Multimodal capabilities: Claude 3 can analyse both text and images, similar to GPT-4. It can process various image formats and compare/contrast multiple images in a single request.
Improved performance: Claude 3 excels at following complex instructions, generating structured outputs, and conversing in multiple languages. It also avoids answering irrelevant questions more often.
Larger context window: Claude 3 considers more information (up to 150,000 words) during conversation, leading to more relevant responses.
Claude 3’s Limitations:
Image processing limitations: Claude 3 cannot identify people in images, struggles with low-quality images, and has difficulty with spatial reasoning tasks.
Limited web access: Unlike GPT-4, Claude 3 cannot search the web for answers, relying solely on pre-loaded data.
Potential for bias and factual errors: Similar to other AI models, Claude 3 is susceptible to bias and might generate inaccurate information.
Future Developments:
Anthropic plans to introduce new features that enhance Claude 3’s capabilities, such as interacting with other systems, coding interactively, and offering advanced agent functionalities.
How to Access Claude 3 in the Pacific Islands
For Pacific Islanders, interested in accessing and utilising Claude 3, here is a link to the Anthropic homepage for more information and access to Claude 3 and its various pricing plans: https://claude.ai/chats. Many consider Claude 3 a high free limit, for personal and non-profit use, where users can ask up to 1000 questions a month.
Source: TechCrunch
*An AI tool was used to add an extra layer to the editing process for this story.
