Last week, IBM’s General Manager of Strategy and Ventures Roger Premo gave a compelling presentation on why IBM may be, at least for the moment, the strongest partner for generative AI implementations. I was at IBM when its was caught flat footed during the client/server and Windows waves.
But we’ve been through recurring waves of technology since. The wave prior to AI was digital transformation where companies were promised an ROI of around 150% and most got somewhere between a -5% and 10% return instead, which was substantially less than what was promised. This is better than most got with client/server technology which didn’t even work at first, video conferencing that no one wanted to use in the 1990s, or big data which was stupid by design (companies built massive data repositories they couldn’t figure out how to mine).
This is often the case with a new technological wave. Sales teams get overly excited and start selling stuff that isn’t close to being cooked, and those of us buying those solutions get screwed.
Why IBM Is Excited (and why you should care)
The reason IBM is excited right now is because, this time, it has the most mature AI platform in watsonx that IBM has been developing for over 20 years, not for just the last 6 months like most everyone else. NVIDIA is the only other company with that kind of deep knowledge, but it’s a parts supplier, not an AI solutions provider.
The entire point of IBM’s pitch was to point out that, unlike its peers, IBM has been working on and deploying AI solutions for decades, not months, and has the most mature set of solutions and processes to assure the result.
IBM’s one lasting trait is that it takes care of its customers. If there is a problem, IBM prioritizes making those customers whole. It’s not the only vendor that does this, but you want that behavior to be a backup in case of unforeseen problems, not because the vendor doesn’t yet know what they are doing. Since AI is so new, most vendors don’t yet know what they are doing with AI.
IBM’s Advice
IBM advises you not to move to deployment when considering AI. First, you should pilot and refine the solution before full deployment. AI has a lot of flavors and variables, so even if the selection is correct in terms of products and services, you’ll still have to fine-tune the result before rolling out the solution.
You want to favor open-source solutions because they benefit from greater developer support and tend to be more secure now because their vulnerabilities are more likely to be caught by third party reviewers (and eventually AIs).
You need to fully understand the large language models (LLMs) you are using, implement best-of-class governance practices, validate, and use tools to de-risk the repositories so you can avoid negative outcomes (like someone else claiming you stole their data, or inherent data corruption or bias).
When you scale the solution, you need to understand how to scale for value. In other words, when AI is applied to different work tasks, the value can range from extremely negative to extremely positive due to a variety of factors you need to fully understand to avoid the negative outcomes that might occur otherwise.
Finally, you need to understand deeply the AI you are using, not just initially, but through the time of the AI’s use because there will be bias introduced, unfairness, model drift along with hallucination problems, and the need for model explainability both to avoid and to forensically analyze problems.
Wrapping Up:
While IBM hasn’t always been at the front end of major tends, it is this time, and is clearly concerned that it isn’t getting credit for having decades more AI experience than any firm other than NVIDIA.
What is sad is that if IBM had retained the huge marketing capability that Louis Gerstner developed (and that Sam Palmisano killed) IBM would be getting that credit today. Microsoft did something similar in the 1990s and its AI efforts are currently hampered as a result. It is a shame that tech companies in general refuse to adequately resource marketing. Even so, IBM’s latest financials highlight its watsonx revenues that doubled quarter-over-quarter which indicates that someone got the memo.
Ironically, marketing is one of the areas where AI could make an enormous difference potentially. In the end, if you want a vendor that isn’t learning about AI but is already extremely experienced with it, there are two choices, IBM and NVIDIA, and the two vendors work together so they aren’t mutually exclusive.
