Stability AI has unveiled its latest image-generation models, the Stable Diffusion 3.5 series, which the company claims will produce more diverse outputs, depicting a wider range of skin tones and features. The series includes three models: the powerful Stable Diffusion 3.5 Large with 8 billion parameters, the faster Large Turbo version, and the Medium model optimized for mobile devices, set to launch on October 29. Stability emphasizes that these models are designed to generate images without extensive prompting and may yield greater output variation. The models are free for non-commercial use, while businesses earning over $1 million annually must obtain an enterprise license. Stability also allows users to request the removal of their data from training sets, with around 80 million images already removed since March 2023. The company remains proactive in addressing copyright concerns and ensuring safe usage of its technology amidst ongoing discussions about AI’s implications in various domains.
Anthropic has launched updated AI models designed to automate mouse clicks and streamline tasks for software developers. The new “computer use” feature allows the AI to perform complex actions like mouse movements and typing commands autonomously. This advancement marks a shift toward AI agents that can execute multi-step processes with minimal human input. Anthropic’s models, including the newly updated 3.5 Haiku and mid-tier Sonnet, are priced based on performance. The company emphasizes the importance of user feedback in refining this feature further. Current safeguards are in place to prevent misuse in areas such as spam and fraud, though Kaplan acknowledges that AI is not infallible.
Anthropic has also updated its Responsible Scaling Policy to manage better the risks associated with increasingly capable AI systems. The new framework emphasizes that as AI capabilities grow, so must safety measures, with the company committing not to train or deploy AI models without implementing adequate safeguards. Anthropic currently assesses its models under four AI safety levels, with all existing models meeting at least the ASL-2 standard.
Microsoft is set to launch its Copilot AI agents next month, allowing customers to automate tasks using a no-code interface. The company claims significant productivity improvements, citing Honeywell’s experience equivalent to adding 187 full-time employees through Copilot usage. In testing, Microsoft reported that sales teams using Copilot achieved a 9.4 percent revenue increase per seller and closed 20 percent more deals. However, experts urge caution regarding these claims, as percentage improvements can be misleading without clear baseline data. As enterprises adopt these agents, whether they will be trusted to make decisions autonomously or merely add to existing alert systems remains to be seen.
Why do we care?
The advancements from Stability AI, Anthropic, and Microsoft signal a broader movement toward the increased adoption of AI tools across various applications—from creative content generation to software development and business operations. These developments represent both opportunities and challenges for MSPs and IT service providers. The need to integrate AI responsibly, demonstrate measurable ROI, and navigate ethical concerns will shape how providers deploy these tools in client environments.
Continuing to know the models and their capabilities is part of the required skills to meet customers’ needs.