The competition in the generative AI assistant space is heating up as Anthropic announced on Tuesday that its new LLM Claude 2 is open for public beta testing. According to the company, Claude 2 can generate code, process large quantities of text, and write long-form responses to prompts.
Anthropic released the LLM’s predecessor, Claude, in March 2023. The company said that it has taken in user feedback to create an improved product, specifically in the fields of coding, math, and reasoning.
“We have heard from our users that Claude is easy to converse with, clearly explains its thinking, is less likely to produce harmful outputs, and has a longer memory,” Anthropic stated.
“Think of Claude as a friendly, enthusiastic colleague or personal assistant who can be instructed in natural language to help you with many tasks,” the company added.
Anthropic said that partners like Jasper, a content strategy AI platform, and Sourcegraph, which works on coding-related AI, have already integrated Claude 2 into their tools.
Currently, beta testing is available to anyone in the US or UK. If you’re outside of these regions, you can easily request access to Claude 2 with the help of a VPN.
How to Access Claude 2 Outside the US and UK
Claude, by design, cannot access the internet, and as with most LLM models, feedback from beta testing will allow the company to improve on its product safety and user experience. People located outside the US or UK interested in trying out Claude 2 can change their IP address with a VPN and sign-up for the beta testing.
Simply follow these steps:
- Get a VPN like NordVPN, which has a vast and reliable network of servers in the US and UK.
- Fire up the app and connect to an American IP address.
- Open your browser and head over to the Claude sign-up page.
Need more instructions? Read our full guide to use Claude from anywhere.
Claude 2 Capable of Processing Technical Documents and Books
Anthropic stated the latest version of its LLM delivers greater performance and safety. It enabled Claude 2 to process longer input and output, which means users can now enter over 75,000 words of text. As a consequence, Claude 2 is capable of digesting and processing hundreds of pages of technical documents and even books, and is able to write longer documents than before.
The company said Claude 2 scores above the 90th percentile on the GRE reading and writing exams, and gets similar results as the average application on the test’s quantitative reasoning section.
The model improved its performance over its predecessor in other tests as well — 76.5% on the Bar Exam’s multiple choice questions, above Claude 1.3’s previous score of 73%. It showed a significant step up in its coding skills, scoring 71.2% on the Codex HumanEva Python test, up from 56%. On grade-school math problems, Claude 2 scored 88% on the on GSM8k — up from 85.2%.
Coming to safety, Anthropic stated its latest model was two times better at giving harmless responses than Claude 1.3.
“We’ve been iterating to improve the underlying safety of Claude 2, so that it is more harmless and harder to prompt to produce offensive or dangerous output. We have an internal red-teaming evaluation that scores our models on a large representative set of harmful prompts, using an automated test while we also regularly check the results manually,” the company wrote.
Shortcomings of Claude 2
Anthropic aims for Claude 2 to benefit regular users in any workplace surrounding. However, it is aware of certain drawbacks with its AI tool. Specifically, the possibility of providing results that can be inappropriate.
The company recommends using the tool for tasks such as summarizing and organizing information. It warns users against using it in situations “where physical or mental health and well-being are involved.”
You can also compare the service with ChatGPT or Google Bard to find which AI model suits you the best.
