-
Anthropic has raised a whopping $7.3 billion in the past year
-
South Korea’s SK Telecom has invested $100 million in Anthropic
-
SK Telecom is part of a telco joint venture working on AI specifically for telcos
One of AI’s hottest startups, Anthropic, announced the third generation of its Claude generative AI chatbot, which competes with OpenAI’s ChatGPT and Google's Bard. And some telcos are already working with Claude to do some telco-specific generative AI innovation.
Founded in 2021 by former OpenAI employees, Anthropic is a generative AI startup based in San Francisco which has ridden the wave of AI mania to raise $7.3 billion in the past year from the likes of Google, Salesforce, Amazon and Menlo Ventures. Generative AI (GenAI) is a type of software that can generate text, images and video based on the data its trained on.
One might ask – why would a company like Google invest in Anthropic, when it’s a direct competitor to Google’s Bard? The answer could be summed up as “hedging its bets.”
Ilya Strebulaev, a finance professor at Stanford, told the New York Times that AI mania is moving so fast investors cannot afford to lose out on the action. “If you miss the winner in the space, you’re kind of out of the game,” she said.
Why is Anthropic of interest to telecom operators?
Because South Korea’s SK Telecom has also invested $100 million in Anthropic to jointly develop a large language model (LLM) optimized for telecom operators. The telco model, which will be a specific variant of Claude, will be overseen by Anthropic co-founder Jared Kaplan, according to Data Center Dynamics.
RELATED: AI joint venture for telcos focuses on customer service to start
In fact, at last week’s MWC conference in Barcelona, SK Telecom, along with Deutsche Telekom, e& Group, Singtel and SoftBank announced plans to establish an AI joint venture, and they held the inaugural meeting of the Global Telco AI Alliance (GTAA).
Through the JV, the five service providers plan to develop large language models specifically tailored to the needs of telcos. And the JV also plans to tap the telco-specific work from Anthropic and SK Telecom.
Claude 3
In terms of Anthropic’s news this week, the company says its Claude 3 family includes three models in ascending order of capability: Haiku, Sonnet and Opus. Each successive model offers increased performance, allowing users to select the optimal balance of intelligence, speed and cost for their specific application.
The company brags that its most intelligent model Opus “exhibits near-human levels of comprehension and fluency on complex tasks, leading the frontier of general intelligence.”
According to Anthropic’s own benchmarks where it compared Opus to OpenAI's GPT-4 and Google's Gemini 1.0 Ultra, Opus beat those competitors in some common tests given to determine an AI's capabilities, such as grade-school math and undergraduate-level knowledge.
The latest version of Claude has also been refined to be more likely to answer prompts and less likely to refuse them.
RELATED: MWC: Microsoft thinks telcos could lead the global AI charge
“The Claude 3 models show a more nuanced understanding of requests, recognize real harm, and refuse to answer harmless prompts much less often,” stated Anthropic. Rather then refusing to answer a prompt, Claude 3 either gives a correct answer or says it doesn’t know the answer. But it aims to avoid an incorrect answer, which the AI industry refers to as “hallucinations.”
The company also appeared to reference plans to implement a technique called grounding to make its AI even more reliable. Within the AI field, grounding is the practice of tying model outputs to specific bits of source material.
“In addition to producing more trustworthy responses, we will soon enable citations in our Claude 3 models so they can point to precise sentences in reference material to verify their answers,” Anthropic added.