IBM Think Conference
Today at its annual THINK conference, IBM revealed many new enhancements to its watsonx platform one year after its launch, as well as planned data and automation features to make AI more open, cost-effective, and flexible for enterprises. IBM CEO Arvind Krishna will discuss the company’s goal to invest in, create, and contribute to the open-source AI community during his opening address.
“Open innovation in AI is IBM’s philosophy. Krishna said they wanted to use open source to do with AI what Linux and OpenShift did. “Open means choice. Open means more eyes on code, brains on issues, and hands on solutions. Competition, innovation, and safety must be balanced for any technology to gain pace and become universal. Open source helps achieve all three.”
IBM published Granite models as open source and created InstructLab, a first-of-its-kind capability, with Red Hat
IBM has open-sourced its most advanced and performant language and code Granite models, demonstrating its commitment to open-source AI. IBM is urging clients, developers, and global experts to build on these capabilities and push AI’s limits in enterprise environments by open sourcing these models.
Granite models, available under Apache 2.0 licences on Hugging Face and GitHub, are known for their quality, transparency, and efficiency. Granite code models have 3B to 34B parameters and base and instruction-following model versions for complicated application modernization, code generation, bug repair, code documentation, repository maintenance, and more. Code models trained on 116 programming languages regularly outperform open-source code LLMs in code-related tasks:
- IBM tested Granite Code models across all model sizes and benchmarks and found that they outperformed open-source code models twice as large.
- IBM found Granite code models perform well on HumanEvalPack, HumanEvalPlus, and reasoning benchmark GSM8K for code synthesis, fixing, explanation, editing, and translation in Python, JavaScript, Java, Go, C++, and Rust.
- IBM Watsonx Code Assistant (WCA) was trained for specialised areas using the 20B parameter Granite base code model. Watsonx Code Assistant for Z helps organisations convert monolithic COBOL systems into IBM Z-optimized services.
- The 20B parameter Granite base code model generates SQL from natural language questions to change structured data and gain insights. IBM led in natural language to SQL, a major industry use case, according to BIRD’s independent leaderboard, which rates models by Execution Accuracy (EX) and Valid Efficiency Score.
IBM and Red Hat announced InstructLab, a groundbreaking LLM open-source innovation platform.
Like open-source software development for decades, InstructLab allows incremental improvements to base models. Developers can use InstructLab to construct models with their own data for their business domains or sectors to understand the direct value of AI, not only model suppliers. Through watsonx.ai and the new Red Hat Enterprise Linux AI (RHEL AI) solution, IBM hopes to use these open-source contributions to deliver value to its clients.
RHEL AI simplifies AI implementation across hybrid infrastructure environments with an enterprise-ready InstructLab, IBM’s open-source Granite models, and the world’s best enterprise Linux platform.
IBM Consulting is also developing a practice to assist clients use InstructLab with their own private data to train purpose-specific AI models that can be scaled to meet an enterprise’s cost and performance goals.
IBM introduces new Watsonx assistants
This new wave of AI innovation might provide $4 trillion in annual economic benefits across industries. IBM’s annual Global AI Adoption Index indicated that 42% of enterprise-scale organisations (> 1,000 people) have adopted AI, but 40% of those investigating or experimenting with AI have yet to deploy their models. The skills gap, data complexity, and, most crucially, trust must be overcome in 2024 for sandbox companies.
IBM is unveiling various improvements and enhancements to its watsonx assistants, as well as a capability in watsonx Orchestrate to allow clients construct AI Assistants across domains, to solve these difficulties.
Watsonx Assistant for Z
The new AI Assistants include watsonx Code Assistant for Enterprise Java Applications (planned availability in October 2024), watsonx Assistant for Z to transform how users interact with the system to quickly transfer knowledge and expertise (planned availability in June 2024), and an expansion of watsonx Code Assistant for Z Service with code explanation to help clients understand and document applications through natural language.
To help organisations and developers meet AI and other mission-critical workloads, IBM is adding NVIDIA L40S and L4 Tensor Core GPUs and support for Red Hat Enterprise Linux AI (RHEL AI) and OpenShift AI. IBM is also leveraging deployable designs for watsonx to expedite AI adoption and empower organisations with security and compliance tools to protect their data and manage compliance rules.
IBM also introduced numerous new and future generative AI-powered data solutions and capabilities to help organisations observe, govern, and optimise their increasingly robust and complex data for AI workloads. Get more information on the IBM Data Product Hub, Data Gate for watsonx, and other updates on watsonx.data.
IBM unveils AI-powered automation vision and capabilities
Company operations are changing with hybrid cloud and AI. The average company manages public and private cloud environments and 1,000 apps with numerous dependencies. Both handle petabytes of data. Automating is no longer an option since generative AI is predicted to drive 1 billion apps by 2028. Businesses will save time, solve problems, and make choices faster.
IBM’s AI-powered automation capabilities will help CIOs evolve from proactive IT management to predictive automation. An enterprise’s infrastructure’s speed, performance, scalability, security, and cost efficiency will depend on AI-powered automation.
Today, IBM’s automation, networking, data, application, and infrastructure management tools enable enterprises manage complex IT infrastructures. Apptio helps technology business managers make data-driven investment decisions by clarifying technology spend and how it produces business value, allowing them to quickly adapt to changing market conditions. Apptio, Instana for automated observability, and Turbonomic for performance optimisation can help clients efficiently allocate resources and control IT spend through enhanced visibility and real-time insights, allowing them to focus more on deploying and scaling AI to drive new innovative initiatives.
IBM recently announced its intent to acquire HashiCorp, which automates multi-cloud and hybrid systems via Terraform, Vault, and other Infrastructure and Security Lifecycle Management tools. HashiCorp helps companies transition to multi-cloud and hybrid cloud systems.
IBM Concert
IBM is previewing IBM Concert, a generative AI-powered tool that will be released in June 2024, at THINK. IBM Concert will be an enterprise’s technology and operational “nerve centre.”
IBM Concert will use watsonx AI to detect, anticipate, and offer solutions across clients’ application portfolios. The new tool integrates into clients’ systems and uses generative AI to generate a complete image of their connected apps utilising data from their cloud infrastructure, source repositories, CI/CD pipelines, and other application management solutions.
Concert informs teams so they can quickly solve issues and prevent them by letting customers minimise superfluous work and expedite others. Concert will first enable application owners, SREs, and IT leaders understand, prevent, and resolve application risk and compliance management challenges.
IBM adds watsonx ecosystem access, third-party models
IBM continues to build a strong ecosystem of partners to offer clients choice and flexibility by bringing third-party models onto watsonx, allowing leading software companies to embed watsonx capabilities into their technologies, and providing IBM Consulting expertise for enterprise business transformation Global generative AI expertise at IBM Consulting has grown to over 50,000 certified practitioners in IBM and strategic partner technologies. Large and small partners help clients adopt and scale personalised AI across their businesses.
- IBM and AWS are integrating Amazon SageMaker and watsonx.governance on AWS. This product gives Amazon SageMaker clients advanced AI governance for predictive and generative machine learning and AI models. AI risk management and compliance are simplified by clients’ ability to govern, monitor, and manage models across platforms.
- Adobe: IBM and Adobe are working on hybrid cloud and AI, integrating Red Hat OpenShift and watsonx to Adobe Experience Platform and considering on-prem and private cloud versions of watsonx.ai and Adobe Acrobat AI Assistant. IBM is also offering Adobe Express assistance to help clients adopt it. These capabilities should arrive in 2H24.
- Meta: IBM released Meta Llama 3, the latest iteration of Meta’s open big language model, on watsonx to let organisations innovate with AI. IBM’s cooperation with Meta to drive open AI innovation continues with Llama 3. Late last year, the two businesses created the AI Alliance, a coalition of prominent industry, startup, university, research, and government organisations with over 100 members and partners.
- Microsoft: IBM is supporting the watsonx AI and data platform on Microsoft Azure and offering it as a customer-managed solution on Azure Red Hat OpenShift (ARO) through IBM and our business partner ecosystem.
- IBM and Mistral AI are forming a strategic partnership to bring their latest commercial models to the watsonx platform, including the leading Mistral Large model, in 2Q24. IBM and Mistral AI are excited to collaborate on open innovation, building on their open-source work.
- Palo Alto Networks: IBM and Palo Alto now offer AI-powered security solutions and many projects to increase client security. Read the news release for details.
- Salesforce: IBM and Salesforce are considering adding the IBM Granite model series to Salesforce Einstein 1 later this year to add new models for AI CRM decision-making.
- SAP: IBM Consulting and SAP are also working to expedite additional customers’ cloud journeys using RISE with SAP to realise the transformative benefits of generative AI for cloud business. This effort builds on IBM and SAP’s Watson AI integration into SAP applications. IBM Granite Model Series is intended to be available throughout SAP’s portfolio of cloud solutions and applications, which are powered by SAP AI Core’s generative AI centre.
- IBM introduced the Saudi Data and Artificial Intelligence Authority (SDAIA) ‘ALLaM’ Arabic model on watsonx, bringing language capabilities like multi-Arabic dialect support.
