What is Salesforce’s new xLAM-1B AI model and why is it significant?
Cover Photo Major News from Salesforce, Zoom, Apple and Cloudflare

Salesforce’s ‘Tiny Giant’ AI Model Outperforms Industry Giants

Salesforce has unveiled a groundbreaking AI model, the xLAM-1B, that outperforms much larger models in function-calling tasks. Despite having just 1 billion parameters, the “Tiny Giant” surpasses industry leaders like OpenAI and Anthropic. The key to xLAM-1B’s success lies in Salesforce’s innovative data curation approach. Their APIGen pipeline generates high-quality, diverse, and verifiable datasets for training AI models in function-calling applications. The research team has made their dataset publicly available, aiming to benefit the wider research community. This could catalyze a new wave of AI development focused on creating hyper-efficient models tailored for specific tasks, rather than one-size-fits-all behemoths.

Zoom Redefines Itself as an “AI-First Collaboration Platform”

Zoom, the $19 billion video-calling giant, is undergoing a strategic shift to reposition itself beyond just video meetings. Graeme Geddes, Zoom’s chief growth officer, recently stated the company’s new aspiration: “We want to be known as an AI-first collaboration platform.” This transformation comes at a crucial time for Zoom, as the demand for remote video conferencing has decreased with businesses implementing return-to-office mandates. To cater to the needs of a hybrid work environment, Zoom introduced its Zoom Workplace suite, which includes virtual whiteboards, guest check-ins, and workspace booking solutions. Zoom’s evolution extends to customer-facing solutions as well, with the company offering chatbot automation services and workflows without video involvement. The company has also acquired the employee engagement platform Workvivo for $272 million, signaling its intent to expand beyond video conferencing.

Apple Secures Observer Role on OpenAI’s Board Amid AI Pact

Tech giant Apple has secured an observer role on the board of OpenAI. According to a Bloomberg News report, Apple’s head of the App Store and former marketing chief, Phil Schiller, has been chosen for the position. The board arrangement is part of a landmark AI agreement announced last month between Apple and OpenAI. The observer role will allow Schiller to attend board meetings, providing him with valuable insights into the decision-making process at OpenAI, although he will not have voting rights or other powers typically held by directors. This development comes on the heels of Apple’s announcement in June, which saw the integration of OpenAI’s ChatGPT technology across its suite of apps, including the virtual assistant Siri. The move is seen as a strategic step by Apple to bolster its AI capabilities and stay competitive in the rapidly evolving technology landscape.

Cloudflare Launches Tool to Combat AI Data Scraping Bots

Cloud service provider Cloudflare has introduced a new, free tool to help website owners combat these unwanted AI bots. The tool aims to detect and block AI bots that are attempting to circumvent standard bot exclusion rules, such as the robots.txt file, to access website content for training AI models. Cloudflare’s analysis of AI bot and crawler traffic has enabled the company to fine-tune its automatic bot detection models, which can identify bots trying to evade detection by mimicking human web browsing behavior.

Cloudflare’s new tool is designed to address the issue of ignoring the robots.txt standard, but its success will depend on the accuracy of its bot detection capabilities. The company has also set up a reporting system for hosts to flag suspected AI bots and crawlers, and it plans to continue manually blacklisting AI bots over time.

Frequently asked questions

Salesforce’s xLAM-1B is a revolutionary AI model nicknamed the “Tiny Giant” that contains only 1 billion parameters but outperforms larger models from companies like OpenAI and Anthropic in function-calling tasks. Its significance lies in its efficiency and innovative data curation approach using the APIGen pipeline, which generates high-quality training datasets. This breakthrough demonstrates that smaller, specialized AI models can be more effective than larger, general-purpose ones.
Zoom is repositioning itself as an “AI-first collaboration platform” rather than just a video conferencing solution. The company has expanded its offerings through the Zoom Workplace suite, which includes virtual whiteboards, guest check-ins, and workspace booking solutions. They’ve also acquired Workvivo for $272 million and introduced chatbot automation services, showing their commitment to evolving beyond traditional video meetings to meet the demands of hybrid work environments.
Salesforce’s APIGen pipeline is a data curation system that generates diverse, verifiable, and high-quality datasets specifically designed for training AI models in function-calling applications. The pipeline has been made publicly available to benefit the broader research community and represents a significant advancement in creating more efficient AI training methodologies.
Apple has secured an observer role on OpenAI’s board, with Phil Schiller, Apple’s head of App Store and former marketing chief, representing the company. While this position allows Schiller to attend board meetings and gain insights into OpenAI’s decision-making process, it doesn’t include voting rights or director powers. This arrangement is part of a larger AI partnership between Apple and OpenAI.
Zoom is responding to the decline in remote work by diversifying its product offerings and focusing on hybrid work solutions. The company has introduced new features through its Zoom Workplace suite, expanded into customer service solutions with chatbot automation, and acquired employee engagement platforms. These changes reflect Zoom’s strategy to remain relevant as businesses implement return-to-office policies.
Cloudflare has launched a free tool that helps website owners detect and block AI bots attempting to scrape their content for AI model training. The tool specifically targets bots that try to bypass standard bot exclusion rules like robots.txt files. It uses advanced bot detection models that can identify bots attempting to mimic human browsing behavior and includes a reporting system for flagging suspected AI bots.
The key differentiator of Salesforce’s xLAM-1B is its ability to achieve superior performance in function-calling tasks despite having only 1 billion parameters, compared to much larger models from competitors. Its success is attributed to Salesforce’s sophisticated data curation approach and the APIGen pipeline, proving that smaller, well-trained models can be more effective than larger ones for specific applications.
Picture of Gor Gasparyan

Gor Gasparyan

Optimizing digital experiences for growth-stage & enterprise brands through research-driven design, automation, and AI