Anthropic on Monday launched Cloud 3, a suite of AI models that it says are the fastest and most powerful yet. The new instruments are called Claude 3 Opus, Sonnet and Haiku.
The most capable of the new models, the Claude 3 Opus, outperformed OpenAI's GPT-4 and Google's Gemini Ultra on industry standard tests, such as college-level knowledge, graduate-level reasoning and basic mathematics, the company said.
This is the first time Anthropic has offered multimedia support. Users can upload images, charts, documents, and other types of unstructured data for analysis and answers.
The other models, the Sonnet and Haiku, are more compact and less expensive than the Opus. Sonnet and Opus are available in 159 countries starting Monday, while Haiku will be released soon, according to Anthropic. The company declined to say how long it took to train Cloud 3 or how much it cost, but said companies like Airtable and Asana helped A/B test the models.
This time last year, Anthropic was seen as a promising AI startup, founded by former OpenAI executives. It has completed Series A and B funding rounds, but has only rolled out the first version of its chatbot without any consumer reach or much fanfare.
Twelve months later, it has become one of the hottest AI startups, with backers including Google, Salesforce and Amazon, and a product that competes directly with ChatGPT in both the enterprise and consumer worlds. Over the past year, the startup has closed five different financing deals, totaling about $7.3 billion.
The field of generative AI has seen a major boom over the past year, with a record $29.1 billion invested across nearly 700 deals in 2023, a more than 260% increase in deal value from the previous year, according to PitchBook. It has become the most popular phrase on corporate earnings calls quarter after quarter. Academics and ethicists have expressed significant concerns about the technology's tendency to spread bias, but despite this, it has quickly made its way into schools, online travel, the medical industry, online advertising and more.
Between 60 to 80 people worked on the basic AI model, while 120 to 150 people worked on its technical aspects, Anthropic co-founder Daniela Amodei told CNBC in an interview. For the latest iteration of the AI model, a team of 30 to 35 people worked on it directly, with about 150 people supporting it, Amodei told CNBC in July.
Anthropic said Claude 3 can summarize up to about 150,000 words, or a large book (think: in the length range of “Moby Dick” or “Harry Potter and the Deathly Hallows”). Its previous version could summarize only 75,000 words. Users can enter large sets of data and request summaries in the form of a memo, letter, or story. In contrast, ChatGPT can handle about 3,000 words.
Amodei also said that Cloud 3 has a better understanding of risks in responses compared to its previous version.
“In our quest for a very innocuous model, Cloud 2 was sometimes overly dismissive,” Amodei told CNBC. “When someone bumps into some of the spicier topics or barriers of trust and safety, Cloud 2 can sometimes be a little reticent in responding to those questions.”
Cloud 3 has a more nuanced understanding of claims, according to Anthropic.
Multimodality, or adding options like image and video capabilities to generative AI, whether you upload it yourself or create it using an AI model, has become one of the most popular use cases in the industry.
“The world is multimedia,” Brad Lightcap, chief operating officer at OpenAI, told CNBC in November. “If you think about the way we as humans deal with the world and interact with it, we see things, we hear things, we say things — the world is much bigger than text. So, for us, it always seemed incomplete that text and code were the individual modalities, the individual interfaces that… “We can have it to see how powerful these models are and what they can do.”
But multimodality and increasingly complex AI models also bring more potential risks. Google recently took its AI image generator, part of its Gemini chatbot, offline after users discovered historical errors and questionable responses, which were widely circulated on social media.
Anthropic's Claude 3 doesn't create images; Instead, it only allows users to upload images and other documents for analysis.
“Of course there is no perfect model, and I think it's very important to say that up front,” Amodei told CNBC. “We've tried hard to make these crossover models to be as capable and safe as possible. And of course there will be places where the model still makes something happen from time to time.”
Clarification: Anthropic explained with CNBC that Cloud 3 can summarize about 150,000 words, not 200,000.
