This Week's Most Read: Small Language Models and Rabbit R1

Here are this week's most popular stories:

1. Small Language Models Gaining Ground at Enterprises

Small language models (SLMs) are becoming more attractive for enterprises to develop and deploy than large language models because they can get more control, such as in fine-tuning for particular domains and data security. They also are cheaper to run.

"We are seeing early adoption of SLMs in enterprises now, especially as hyperscalers like AWS and Azure are providing access to these models as hosted APIs,” said Pushpraj Shukla, senior vice president of engineering and head of AI/ML at SymphonyAI.

An SLM is generally five to 10 times smaller than LLMs and are open source projects. The smaller size means much lower energy consumption. They can also be hosted on a single GPU. This is a major benefit given the shortage of these chipsets and the steep cost of compute.

Despite their reduced size, SLMs demonstrate capabilities that are remarkably close to LLMs in various NLU tasks. This is especially the case when they are effectively fine-tuned (or retrained) for specialized use cases, say health care or coding. The process can take minutes to several hours, compared to tens of hours to a few days for LLMs. To get effective results with an SLM, the dataset often should have several hundred thousand examples.

Read more

2. Exclusive: Salesforce EVP on Deploying Einstein and CRM AI

Rahul Auradkar, executive vice president and general manager of unified data services and Einstein at Salesforce, recently joined the AI Business Podcast to discuss the Einstein platform and other key AI initiatives.

Listen to the podcast or read the transcript.

3. Rabbit R1, a Novel AI Agent Device, Is Selling Out

Welcome to the new world of AI agents: assistants that not only understand what you want to do but execute it for you. A recent AI agent entrant that has been the talk of the AI community is rabbit’s R1 device. It is a standalone, $199 device that connects to your mobile apps and executes tasks for you − using a proprietary new language model and without tapping into any APIs.

The r1 takes instructions in natural language. For example, ask it to call an Uber for a family of four and it will do so. The user confirms the transaction. The device is also teachable: in a demo, it learned how to generate images after the user tapped into Midjourney to show it how.

As of Jan. 18, the fifth batch of 10,000 r1 devices has sold out, according to a company tweet. Pre-orders for the sixth batch of 50,000 devices is open at rabbit.tech, with delivery dates in June and July.

One fan is Microsoft CEO Satya Nadella. "I thought the demo of the rabbit OS and the device was fantastic,” he told Bloomberg. “After (Steve) Job’s launch of iPhone, (it was) probably one of the most impressive presentations I have seen of capturing the vision of what is possible going forward for what is an agent-centric operating system and interface.”

4. AI and Cybersecurity: Guard Against ‘Poisoning’ Attacks

AI systems are vulnerable to bad actors infusing them with bad data, a technique known as ‘poisoning attacks,’ according to the co-author of a new U.S. government study.

The National Institute of Standards and Technology study analyzed cyber threats to AI systems amid rising concerns over the safety and reliability of generative AI as the 2024 election cycle heats up.

“Most of these attacks are fairly easy to mount and require minimum knowledge of the AI system and limited adversarial capabilities,” said study co-author Alina Oprea, who is a Northeastern University professor. “Poisoning attacks, for example, can be mounted by controlling a few dozen training samples, which would be a very small percentage of the entire training set.”

The integrity of the data used for training these AI systems is a significant concern. Often sourced from websites and user interactions, the data is susceptible to manipulation by malicious entities.

Read more

5. AI Startup Roundup: OpenAI Rival Cohere in Talks to Raise $1 Billion

OpenAI rival Cohere is in talks to raise up to $1 billion in funding, according to a report by FT.

It will be the most money ever to come to the Canadian startup as the AI race heats up. Cohere had four funding raises thus far, with the last one in June 2023 putting the startup’s valuation at $2.2 billion. Backers back then were Nvidia and Oracle as well as VC firms Index Ventures and Inovia Capital.

Former Google scientists Aidan Gomez and Nick Frosst founded Cohere, along with Ivan Zhang.

Gomez co-authored the famous seminal paper on Transformers that has revolutionized large language models. Frosst worked at Google Brain under Turing award winner Geoffrey Hinton’s Toronto team.

Cohere develops large language models for enterprises to build custom applications such as AI chatbots, using their own data. In contrast, OpenAI’s large language models are meant for wider use for consumers and businesses alike.

Cohere is one of AI Business’ seven companies to watch in 2024.Read more