IBM has now made an exciting announcement about the launch of the initial AI models as part of the watsonx Granite series. These models, rooted in generative AI technology, are crafted to carry out language- and code-related tasks.
The variants of these models are adjusted to match diverse business requirements, since they are all architectured on the fundamentals of the decoder-only mechanism. These AI models can offer scalability across various functions, such as creating personalized responses from organization knowledge repositories, condensing massive content like agreements or call logs, and gathering insights alongside the classification of data like customer sentiment analysis.
Recognizing that one size does not fit all, IBM has also made arrangements for the integration of third-party models like Meta's Llama 2-chat, a model that enjoys up to 70 billion parameters and others from the Hugging Face community.
Affirming the strategic importance of AI, Dinesh Nirmal, Senior Vice President of Products at IBM Software, explained: In the current landscape of AI innovation, businesses destined for success align themselves with AI technologies embodying scalable success and possessing strong built-in frameworks and principles for responsible use. Importantly, he drew attention to IBM's commitment to standing behind WatsonX models and the release of the Granite model series as clear reflections of IBM's holistic model lifecycle management within the watsonx AI and data platform. This provides businesses with a competitive edge with cutting-edge AI custom-designed for their unique needs.
IBM has invested time and resources into the development of fundamental AI models by training them on diverse datasets across five domains: internet, academia, code, legal, and finance. These models have been carefully evaluated by IBM to ensure their suitability for business applications. The rigour involved in the training data screening mechanism to cleanse the objectionable content, in conjunction with the benchmarking activities carried out against both external and internal models, underscores IBM's commitment to the project.
Moreover, IBM's objectives include ensuring responsible AI deployments by addressing governance, privacy, risk assessments and bias mitigation. They achieve this by leveraging their own AI and data model lifecycle governance processes, thereby managing and minimising risk for clients through the WatsonX AI and data platform.
Re-iterating IBM's commitment to trusted AI workflows, it revealed plans to release watsonx.governance, an AI governance toolkit, later in the year. Alongside this, IBM confirmed that its standard intellectual property protections would naturally be extended to these AI models.
As developments from giants like IBM unfold, companies such as AppMaster are also shaping the future of tech, harnessing the power of no-code tools to create feature-rich backend, web, and mobile apps, with more to look forward to in the coming months.
