American AI company Anthropic has launched its latest artificial intelligence model, Claude 2. As a highly trained conversational chatbot, Claude 2 rivals OpenAI’s global sensation, ChatGPT. So, what is Claude 2 and what can it do?
As one of the first constitutional AI chatbots, not only has Claude 2 got improved performance and longer responses, but it can make judgments about the content it is creating. This puts it a step ahead of many large language models in a way that is both impressive and terrifying. This article will explain what this latest chatbot is and some of the features that set Claude 2 apart from the crowd.
Let’s get into it.
Essential AI Tools
Content Guardian – AI Content Checker – One-click, Eight Checks
Jasper AI
WordAI
Copy.ai
Writesonic
What is Claude 2?
Described by Anthropic as a “friendly, enthusiastic colleague or personal assistant”, Claude 2 is a natural language processing chatbot. It sees improvements from the previous model in reasoning, math, and coding. As previously mentioned, Claude AI is a constitutional chatbot. Essentially, it has been trained to make judgments based on a set of principles taken from documents including the 1948 UN Declaration and Apple’s terms of service. This makes sure it expands to issues in the digital domain. It has complex choice-making capabilities and scored 76.5% on the multiple-choice section of the bar exam. While it hasn’t quite had the publicity of the big chatbots ChatGPT, Google Bard, or Bing Chat, it’s shaping up to be an impressive rival in performance metrics.
Claude 2 is available as an API. Anyone in the US and the UK can now start using it on Anthropic’s new public-facing beta website.
What are some of the features of Claude 2?
Claude 2 has a larger context window than other models and is able to explain its thinking. It can also perform impressive coding and mathematics. Here are some of its main features.
- Improved coding skills, with Claude 2 scoring 71.2% on a Python coding test.
- Increased input and output. Users can input 100k tokens in each prompt (about 75,000 words!), which will get analyzed by Claude in mere minutes.
- Better reasoning and rationality. In the GRE reading and writing exams, Claude was in the 90th percentile and scored similarly to the median applicant on quantitative reasoning.
- Natural language skills. Is a multilinguistic model, can process complex language prompts, and quickly create summaries from large text files
Anthropic is currently working on making Claude 2 more globally available. At the time of writing, the beta website in which users can try out the model is only available in the US and the UK. In terms of pricing, the chatbot beta is currently free. However, usage of the platform may be limited for unpaid users, and Claude 2 prompts are available to purchase for $11.2 / million tokens.
Claude Instant
If you’re on a budget, or don’t actually need the full spectrum of features of the Claude 2 AI model, you could use Claude Instant. Anthropic’s lightweight chatbot launch offers you a limited range of the features of Claude 2 for a cheaper price. Instant costs $1.63 per million tokens to prompt, and $5.51 per million tokens of completed character output – with a context window of 100k tokens.
Claude Instant recently got an impressive update from 1.1 to 1.2. The update incorporates some of the strengths of the Claude 2 model, including improvements in key areas like math, coding, reasoning, and safety. It’s currently available as both a chat interface as well as an API.
Final thoughts
Claude 2 is certainly an impressive AI startup, with the potential to shift ChatGPT and GPT-4 from its spot as the top language model. The reasoning capabilities of this AI assistant demonstrate impressive advances in artificial intelligence’s understanding of human morality, and the Claude API could become a hugely popular LLM. However, Anthropic, started by former senior members of the OpenAI team Daniela and Dario Amodei, claims to be an ethically driven business with AI safety at its core. If you’re interested in AI, we have articles on the latest models including Google Bard and GTP-4.
