New Hardware for Smart Software from IBM
November 2, 2022
IBM is getting into the AI hardware acceleration game with its new Artificial Intelligence Unit (AIU), we learn from VentureBeat‘s piece, “IBM Announces System-On-Chip AI Hardware.” Each AIU holds 32 cores similar to the Telum chip’s AI core. Rather than a CPU or GPU, the new component is an application-specific integrated circuit (ASIC) designed with AI in mind. This allows it to perform tasks not part of many AI accelerators, we’re told, like the ability to virtualize AI acceleration services. We are assured it is compatible with “the vast majority” of software commonly used by data scientists.
So far so good, but we noticed something in a passage tucked at the end of the write-up—it almost seems results are merely close enough for horseshoes and hand grenades. In order to work faster, the AIU practices “approximate computing.” Kerner tells us:
“Approximate computing is really the recognition that AI is not 100% correct,’ Leland Chang, principal research staff member and senior manager, AI hardware, at IBM Research, told VentureBeat. Chang explained that AI often works by recognizing a pattern and could well be just 99% accurate, meaning that 1% of results are incorrect. The concept of approximate computing is the recognition that within the AI algorithm it is possible to cut some corners. While Chang admitted that this can reduce precision, he explained that if information is lost in the right places, it doesn’t affect the result — which, more often than not, will still be 99% correct. ‘Approximate computing … is simply recognizing that it doesn’t have to be 100% exact,’ Chang said. ‘You’re losing some information, but you’re losing in places where it doesn’t matter.'”
You don’t say. Can we get a guarantee of that? Who makes the electronic components? Oh, right. Bad question.
Cynthia Murrell, November 2, 2022
IBM Data Governance Tools
October 21, 2022
Confused about data governance? Just rely on IBM. That is our takeaway from a write-up at TechRepublic, “An Overview of IBM Data Governance Solutions.” Author Aminu Abdullahi begins by describing IBM’s top data tools, though whether “top” here means most popular or most heavily promoted is unknown. First up is Cloud Pak, a cloud-based AI platform made to gather and analyze data from multiple sources. OpenPages both guides users in protecting sensitive data and manages compliance issues. To wrest BI insights from data, users can turn to InfoSphere Optim. Then, of course, there is everything Watson. The post explains the framework that pulls it all together:
“As an organization grows, it’s important to have a plan to protect and manage data. The IBM data governance framework is a set of best practices that helps businesses create an overarching strategy for managing the life cycle of their data. IBM’s data governance practice framework includes four types of control:
- Ensure: Controls for guiding work.
- Assure: Controls for doing work.
- Insure: Controls for operating.
- Reassure: Controls for continuity.
These controls allow companies to identify, protect, manage, monitor and report on their data. They do this by working with their business leaders, functional heads and IT teams across the organization to create unified standards for how companies should use information from creation through disposal. For example, the Identify phase will help establish roles and responsibilities for stakeholders within the organization; Protect will provide guidelines for how to store all types of data securely; Manage can help ensure high-quality information; Monitor can give insight into what’s happening with information assets. Finally, Report covers tools that generate comprehensive reports on all aspects of data management. The framework helps build an environment where accountability and responsibility are clear across the enterprise.”
So IBM is a one-stop shop for responsible and profitable data management, if you will. The post concludes by noting these tools have received rave reviews from current users. We wonder, though, how many of those users have any basis for comparison. We ask, “Can’t Watson do this?”
Cynthia Murrell, October 21, 2022
The Push for Synthetic Data: What about Poisoning and Bias? Not to Worry
October 6, 2022
Do you worry about data poisoning, use of crafted data strings to cause numerical recipes to output craziness, and weaponized information shaped by a disaffected MBA big data developer sloshing with DynaPep?
No. Good. Enjoy the outputs.
Yes. Too bad. You lose.
For a rah rah, it’s sunny in Slough look at synthetic data, read “Synthetic Data Is the Safe, Low-Cost Alternative to Real Data That We Need.”
The sub title is:
A new solution for data hungry AIs
And the sub sub title is:
Content provided by IBM and TNW.
Let’s check out what this IBM content marketing write up says:
One example is Task2Sim, an AI model built by the MIT-IBM Watson AI Lab that creates synthetic data for training classifiers. Rather than teaching the classifier to recognize one object at a time, the model creates images that can be used to teach multiple tasks. The scalability of this type of model makes collecting data less time consuming and less expensive for data hungry businesses.
What are the downsides of synthetic data? Downsides? Don’t be silly:
Synthetic data, however it is produced, offers a number of very concrete advantages over using real world data. First of all, it’s easier to collect way more of it, because you don’t have to rely on humans creating it. Second, the synthetic data comes perfectly labeled, so there’s no need to rely on labor intensive data centers to (sometimes incorrectly) label data. Third, it can protect privacy and copyright, as the data is, well, synthetic. And finally, and perhaps most importantly, it can reduce biased outcomes.
There is one, very small, almost miniscule issue stated in the write up; to wit:
As you might suspect, the big question regarding synthetic data is around the so-called fidelity — or how closely it matches real-world data. The jury is still out on this, but research seems to show that combining synthetic data with real data gives statistically sound results. This year, researchers from MIT and the MIT-IBM AI Watson Lab showed that an image classifier that was pretrained on synthetic data in combination with real data, performed as well as an image classifier trained exclusively on real data.
I loved the “seems to show” phrase I put in bold face. Seems is such a great verb. It “seems” almost accurate.
But what about that disaffected MBA developer fiddling with thresholds?
I know the answer to this question, “That will never happen.”
Okay, I am convinced. You know the “we need” thing.
Stephen E Arnold, October 6, 2022
IBM Power10 Rah Rah: One Concerning Statement
September 12, 2022
IBM is back in the marketing game. Everyone wants a Power10 computer in a mobile phone or a MacBook Air form factor. Am I right! Yes.
The article “IBM Power10 Shreds Ice Lake Xeons for Transaction Processing.” This is a big iron made less big. The article points out use cases for those AIX users. Plus there are references to notable big iron outfits like Oakridge and Lawrence Livermore Labs, both really common computing environments like those in the local Coca-Cola distributor’s office or the regional garbage outfit’s offices in three cities.
The charts are phenomenal. Here’s an example. Look at how the blue bar is lower than the gray bar. And the power savings and the thermal data? You know what air conditioners are for as well as those nifty Caterpillar generators in the parking lot are for, don’t you?
Very encouraging.
But…
I noticed one sentence which gave me pause; to wit:
IBM will, of course, make some competitive wins, mostly in emerging markets (and in some cases as Inspur sells iron in China), and it will also win some deals for new kinds of workloads like MongoDB, EnterpriseDB, or Redis.
With the export restrictions imposed by the US on China, will the Power10 find its way to the Middle Kingdom? The use cases for Power10 at US national laboratories may exist in a country wrestling with some real estate issues. Can the Power10 help with the land and construction challenges? What about Chinese academics-only, please research outfits?
In the midst of a PR type content marketing article, I found the reference to China interesting. Will anyone else?
Stephen E Arnold, September 12, 2022
The Home of Dinobabies Knows How to Eliminate AI Bias
August 26, 2022
It is common knowledge in tech and the news media that AI training datasets are flawed. These datasets are unfortunately prone to teaching AI how to be “racist” and “sexist.” AI are computer programs, so they are not intentionally biased. The datasets that teach them how to work are flawed, because they contain incorrect information about women and dark-skinned people. The solution is to build new datasets, but it is difficult to find hoards of large, unpolluted information. MIT News explains there is a possible solution in the article: “A Technique To Improve Both Fairness And Accuracy In Artificial Intelligence.”
Researchers already know that AI contain mistakes so they use selective regressions to estimate the confidence level for predictions. If the predictions are too low, then the AI rejects them. MIT researchers and MIT-IBM Watson AI Lab discovered what we already know: women and ethnic minorities are not accurately represented in the data even with selective regression. The MIT researchers designed two algorithms to fix the bias:
“One algorithm guarantees that the features the model uses to make predictions contain all information about the sensitive attributes in the dataset, such as race and sex, that is relevant to the target variable of interest. Sensitive attributes are features that may not be used for decisions, often due to laws or organizational policies. The second algorithm employs a calibration technique to ensure the model makes the same prediction for an input, regardless of whether any sensitive attributes are added to that input.”
The algorithms worked to reduce disparities in test cases.
It is too bad that datasets are biased, because it does not paint an accurate representation of people and researchers need to fix the disparities. It is even more unfortunate locating clean datasets and that the Internet cannot be used, because of all the junk created by trolls.
Whitney Grace, August 26, 2022
IBM Smart Software and Technology: Will There Be a Double Fault?
July 9, 2022
It has been a few years since Wimbledon started using AI to engage fans and the media. The longstanding partnership between IBM and the venerable All England Lawn Tennis Club captured the Best Fan Engagement by a Brand trophy at the 2022 Sports Technology Awards. The “IBM Power Index with Watson,” “IBM Match Insights with Watson,” and “Personalized Recommendations and Highlights Reels” were their winners. Maybe Watson has finally found its niche. We learn what changes are in store this season in the company’s press release, “IBM Reveals New AI and Cloud Powered Fan Experiences for Wimbledon 2022.” The write-up specifies:
“New features for 2022 include:
* ‘Win Factors’ brings enhanced explainability to ‘Match Insights’: Building on the existing Match Insights feature of the Wimbledon app and Wimbledon.com, IBM is providing an additional level of explainability into what factors are being analyzed by the AI system to determine match insights and predictions. Win Factors will provide fans with an increased understanding of the elements affecting player performance, such as the IBM Power Index, court surface, ATP/WTA rankings, head-to-head, ratio of games won, net of sets won, recent performance, yearly success, and media punditry.
* ‘Have Your Say’ with a new interactive fan predictions feature: For the first time, users can register their own predictions for match outcomes on the Wimbledon app and Wimbledon.com, through the Have Your Say feature. They can then compare their prediction with the aggregated predictions of other fans and the AI-powered Likelihood to Win predictions generated by IBM.”
The “digital fan experiences” use a combination of on-premises and cloud systems. Developers have trained the machine-learning models on data from prior matches using Watson Studio and Watson Discovery. See the press release for more specifics on each feature.
Cynthia Murrell, July 9, 2022
IBM Seeks to Avoid Groundhog Day in AI/ML
July 8, 2022
How do you deliver the killer AI/ML system? Via news releases and PR perhaps?
The Next Web claims that, “IBM’s Human-Centered Approach Is The Only Big Tech Blueprint AI Startups Should Follow.” Author Tristan Greene reminds readers that IBM’s initials stand for International Business Machines and he met the company’s first chief AI officer Seth Dobrin. Dobrin said IBM would never focus on consumer AI, i.e. virtual assistants and selfie apps.
Dobrin also stated that IBM’s goal is to create AI models that improve human life and provide value for its clients and partners. It is apparently not hard to do if you care about how individuals will be affected by monetized models. He compared these models to toys:
“During a discussion with the Financial Times’ Tim Bradshaw during the conference, Dobrin used the example of large-parameter models such as GPT-3 and DALL-E 2 as a way to describe IBM’s approach.
He described those models as “toys,” and for good reason: they’re fun to play with, but they’re ultimately not very useful. They’re prone to unpredictability in the form of nonsense, hate speech, and the potential to output private personal information. This makes them dangerous to deploy outside of laboratories.
However, Dobrin told Bradshaw and the audience that IBM was also working on a similar system. He referred to these agents as “foundational models,” meaning they can be used for multiple applications once developed and trained.”
IBM takes a human approach to its projects. Instead of feeding its AI datasets that could contain offensive information, IBM checks the data first before experimenting. That way the AI is already compliance ready and there will not be any bugs to work out later (at least the prejudice type). IBM is also focused on outcomes, not speculation, which is not how the tech giants work.
IBM wants to withstand an AI winter that could come after the fancy lights, parlor tricks, and flashy PR campaigns are in the past. Human-centered AI technologies, as Dobrin believes, will last longer and provide better services. IBM is also dedicated to sustainability.
IBM is green and wants to create better products and services before launch? It sounds better than most, but can they deliver?
Whitney Grace, July 8, 2022
Ahoy, Captain Watson, Will We Make It This Time, Arrrghh
June 10, 2022
Not one to let repeated failures get in its way, marine research non-profit ProMare has once again sent its Mayflower Autonomous Ship across the open ocean with Watson at the helm. The Register reports, “IBM-Powered Mayflower Robo-Ship Once Again Tries to Cross Atlantic.” When the project first embarked in 2020, we wondered whether it might fall victim to hackers. As it turns out, that attempt was foiled by a more basic issue—a mechanical fault with its generator. As advanced as it is, Watson cannot yet wield a physical wrench. A minor electrical glitch halted the more recent crossing attempt, launched this past April 28, two weeks in. That issue was quickly fixed and the ship set on its way once again. Reporter Katyanna Quach writes:
“‘As of 0900 BST May 20, MAS was back underway with its transatlantic crossing,’ the IBM spokesperson said. It is aiming to complete the remaining 2,225-mile voyage in 16 days. Now, nearly a week into resuming its journey, the ship has made it to its furthest distance yet, a little over halfway to America.”
Well, that was a couple weeks ago. As of this writing, the MAS has been diverted to Nova Scotia to address yet another electrical issue. This team is nothing if not persistent. Quach goes on to give us a few details about the tech involved:
“The Mayflower’s AI software runs on four computers containing Intel processors, six nVidia Jetson AGX Xavier GPUs, two nVidia Jetson Xavier NX boards, and a few other chips. Live camera footage streaming from a webcam onboard the ship is back up online for viewers to follow. ‘We’ve made lots of improvements – the computer vision system has been significantly improved through at-sea testing, and similarly the data fusion algorithms are functioning better and better with every deployment and have greatly improved over the course of the past year,’ Brett Phaneuf, co-director of the Mayflower project … told The Register in a statement. ‘We’ve also improved many mechanical systems, particularly the air intake and exhaust for the generator on the hybrid drive line – and we’ve reduced power consumption significantly as well, over the past year, through applied research, testing and trials, and we’ve made the boat more robust in general.’”
Not quite robust enough, it seems. Not yet. It looks like ProMare is determined to press Watson past its limits. Will the persistent little ship finally make it to its destination? Curious readers can follow MAS’ progress here.
Cynthia Murrell, June 10, 2022
Can Kyndryl Drill IBM and Strike Gold? But Who Drilled Whom?
June 6, 2022
I found it amusing that Big Blue found itself on the wrong side of what I call a deal poaching allegation. The details of the BMC and IBM services for AT&T is interesting. However, a knock on effect of that $1.6 billion dollar settlement has put the estimable IBM spin out Kyndryl in the spotlight. I thought the name “Kyndryl” was one of those pharma products tailored to old people who watched cable news talking heads. Was I wrong? Absolutely. Kyndryl is IBM’s managed services business. The idea for that was that big companies did not want to deal with full time equivalents who kept an organization’s servers chugging along. Let IBM do it was a good business until the sharpies at Amazon, Google, and Microsoft, among others, figured out how to package the cloud to chew into Big Blue’s revenues.
Gulp down a Wal-DRYL and check out “Kyndryl Shares Swoon on Fear It Faces Huge Liability in Lawsuit Against IBM.” The article explains:
Spun out of IBM (ticker: IBM) last year, Kyndryl (KD) is basically IBM’s old managed IT services business. It is a gigantic company with around $18 billion in sales and a workforce of about 90,000 people. But Kyndryl is shrinking at the top line, pays no dividend, and is having trouble finding a constituency among investors. The stock is down more than 50% since the spinoff was completed in November. And now the two companies find themselves on the opposite sides of a legal mess that poses considerable risks for Kyndryl.
What I find interesting is that the incident strikes me as one part of IBM is fighting another part of IBM. Where does the customer fit into this crashing of brilliantly managed entities? I know. Let’s ask Watson. No, let’s check out M-Dryl. Seems less financially risky.
Stephen E Arnold, June 6, 2022
IBM: Fueling the Quantum Computing PR Push
May 25, 2022
I wonder if your mom wants a quantum computer in her tablet. Who wants a quantum computer? Many people, and IBM wants to deliver. The issue is that the fungible quantum computer is a bit of a specialty item. IBM wants to be perceived as the Big Dog in the sector.
“IBM Plans to Deliver 4 000+ Qubit System” explains that IBM will produce a quantum computing Big Blue Great Dane of a system. (Did you know that there are more than a dozen dogs which have blue coats? In addiiton to the Great Dane, there is the blue Chihuahua and the blue Lacy.) The important word to me is “plans.” The 4,000 qubit giant is not ready for the quantum computer market yet. But it is coming. Soon. And the road map will be updated in 2023.
The write up says:
IBM has announced the expansion of its roadmap for achieving large-scale, practical quantum computing. This roadmap details plans for new modular architectures and networking that will allow IBM quantum systems to have larger qubit-counts – up to hundreds of thousands of qubits.
Note that the numerical leap is from 4,000 qubits to hundreds of thousands of qubits.
And IBM’s innovations are not Google-style pronouncements of quantum supremacy. IBM, according to the article:
will leverage three pillars: robust and scalable quantum hardware; cutting-edge quantum software to orchestrate and enable accessible and powerful quantum programs; and a broad global ecosystem of quantum-ready organizations and communities.
In addition, IBM will roll out IBM Condor, not a Big Blue dog but a condor, which is a Big Bird. I noted this statement:
On the hardware front, IBM intends to introduce IBM Condor, the world’s first universal quantum processor with over 1 000 qubits.
The Big Blue Great Dane and the Big Bird are part of the road map. Is IBM Watson available to answer a question about this forward leaning quantum computing PR announcement about “plans”?
Stephen E Arnold, May 25, 2022
-
- Subscribe to Beyond Search
Feature archive
News archive
-
