To place Claude AI accurately within the LLM universe, it’s worth noting that Anthropic, its parent company, was founded in 2021 by two ex-OpenAI employees who left due to their concerns about safety and ethics inherent in the early models of GPT. It is a business that exists to make a positive impact on society (a US-based public-benefits corporation).
It has privacy and freedom at the heart of its algorithm, being guided by Anthropics Constitution. It has a purpose, rather than the vague human-centric guidance that other models such as GPT are guided by.
Claude’s main selling point is the amount of data it can interpret. There are options to upload text or PDFs up to 10MB, or 100,000 words. Given the limits of other LLMs in this area it makes Claude the go-to for anyone looking to interpret PDFs or even large book excerpts (legally obtained, of course) and discuss them with an LLM.
This capability is also available for free, though there are limits on usage. This is still a step-up from GPT that doesn’t allow any attachment uploads on a free account.
Aside from that, this is a no-frills experience, with no other plugins available and no ability to interpret or produce images. However, this is likely the way that generative AI models will progress from where they are now (as of the end of 2023). A model that specialises and does one things very well will be more valuable than a general model that doesn’t excel at anything.
In keeping with its ethical grounding, Claude is not connected to the internet, allowing more control over its content. However, it is regularly trained on updated information so isn’t as stuck in the past as GPT 3.5.
