In a groundbreaking move, Anthropic, fueled by Google’s support and substantial venture capital, introduces Claude 3, its latest GenAI technology. This family of models, including Haiku, Sonnet, and the powerful Opus, claims superiority over OpenAI’s GPT-4 in terms of performance.
Anthropic’s Claude 3 marks a significant leap as its first multimodal GenAI, proficient in analyzing both text and images, a feature akin to GPT-4 and Gemini. Able to process diverse media such as photos, charts, graphs, and technical diagrams from various document types, Claude 3 sets itself apart.
However, there are some constraints. Claude 3 refrains from identifying people in images, emphasizing ethical considerations. It may also stumble with low-quality images and struggle in tasks involving spatial reasoning and object counting. Notably, it won’t generate artwork – its focus is strictly on image analysis.
Anthropic asserts that Claude 3 excels in following multi-step instructions, producing structured output, and conversing in languages beyond English. The model boasts a nuanced understanding of requests, reducing instances of refusing to answer questions. Users can soon expect answers with cited sources for verification.
The upgrade includes an expanded context for Claude 3, with a 200,000-token context window initially, reaching up to 1 million tokens for select customers. This enhancement ensures better comprehension of the narrative flow and richer, more contextually relevant responses.
While Claude 3 isn’t flawless, acknowledging issues of bias and hallucinations, Anthropic promises frequent updates in the coming months. The model is currently restricted to pre-August 2023 data, unable to search the web.
Anthropic offers Opus, Sonnet, and soon Haiku at varying prices, making Claude 3 accessible to a broad audience. Opus, the most powerful, is priced at $15 per million input tokens and $75 per million output tokens.
Anthropic’s ambitions extend beyond just improving models. Their goal is to create an AI algorithm for “AI self-teaching,” envisioning virtual assistants capable of diverse tasks like answering emails, conducting research, and generating art and books. Claude 3’s future updates aim to enhance its capabilities, allowing it to interact with other systems, code interactively, and deliver advanced agentic functions.
The pricing breakdown for Claude 3 models:
- Opus: $15 per million input tokens, $75 per million output tokens
- Sonnet: $3 per million input tokens, $15 per million output tokens
- Haiku: $0.25 per million input tokens, $1.25 per million output tokens
Anthropic emphasizes that model intelligence has room for growth, promising continuous enhancements to Claude 3. The company’s commitment is evident in its goal to raise to $5 billion in the next 12 months, showcasing strong support from major players like Google and Amazon.
Intriguingly, Anthropic’s technique for training GenAI, known as “constitutional AI,” aims to align AI behavior with human intentions. This approach involves models responding to questions and performing tasks based on a set of guiding principles, ensuring transparency and adjustability.
As Anthropic navigates the competitive landscape, its journey promises innovation and evolution. Whether it’s image generation or advanced agentic capabilities, Claude 3 hints at a future where AI seamlessly integrates with complex tasks. With a commitment to understanding and addressing ethical concerns, Anthropic aims to shape the future of AI.
