AI & ML

How Much of Vishal Sikka’s Explanations on Artificial Intelligence is Complete BS?

Executive Summary

  • We have known Vishal Sikka to be a person who provides highly inaccurate information on whatever topic he discusses.
  • Vishal Sikka has recently been discussing AI for his new startup. How much of what Vishal’ describes is real?

Introduction

Back in 2015, and when he was head of the outsourcing firm Infosys. Infosys, like the other major consulting companies, were faking their AI capabilities at the time. Vishal Sikka, who has a P.h.D with a dissertation in Artificial Intelligence, never really worked much in the field as we cover in the article Is Oracle Board Member Vishal Sikka a World Expert on AI?. People who are the CEOs of companies with hundreds of thousands of employees don’t get their hands dirty with actual technical work. Still, part of Vishal Sikka’s job is to do PR for Infosys (at the time). We found curious quotes from his 2015 post Open AI For All.

Vishal is Concerned for the “Interests of Humanity?”

This portion of his blog post jumped out to us.

My friend and teacher Alan Kay once referred to Sam Altman as a “builder of civilizations”.  When Sam, a wise man who is but 30 years old, was thinking about the idea of building an open ecosystem for, among other endeavors, AI, Alan and I shared our ideas with him and our experiences.  Sam asked me if I would be ok with the fact that such an endeavor would be untethered and would produce results generally in the greater interests of humanity,(emphasis added) and he was somewhat surprised by my reaction, that indeed I would only support this venture if such an openness was a fundamental requirement!  In all my experience with corporate research teams, I found a continual struggle for the teams to find relevance with the work in the “here and now”, usually knowing that this unnecessary and premature seeking of relevance not only blinds us to those opportunities that can shift our paradigms, it defeats the point of research.

The deception here is quite apparent.

Vishal Sikka is an enormously overpaid executive who, throughout his career, has provided an enormous amount of false information out to the marketplace — for whatever company he was working for. The idea that Vishal Sikka cares about..

“the interests of humanity”

..is not believable.

Vishal’s Deep Thoughts About Newton and Other Scientific Giants?

In this quote, Vishal again tries to communicate his “big thoughts.”

He shares the view that cooperation helps dramatically improve our lot, helps create a foundation for much larger value creation than any isolated “feudal” system can.  Indeed, endeavors such as agriculture, and science, show that when we share, we improve all of us.  As Newton once said, “if I have seen further, it is by standing on the shoulders of giants”.

This is a pattern we have noticed with Vishal Sikka, in that he pitches what high minded concepts are, and then refers to great scientists. Mostly its a waste of time, and his references seem shoehorned into this prose.

If we look at his comment around “feudalism,” it is incongruous for Vishal to write about elite institutions like feudalism. Especially when considering that Infosys was an is a bottom-feeding body shop where workers are poorly treated, and the elites in Infosys like Vishal are extraordinarily overpaid for the contributions. Vishal was born in India, and again India is one of the least income equal countries in the world. The Indians that migrate to the US come from the upper parts of society and usually bring elite attitudes with them.

Great Leaders….Like Infosys?

After referring to Newton, Vishal then puts Infosys, the bottom of the barrel outsourcing firm, with AWS in AI?

I am really excited that Sam, Elon Musk, and others — including Reid Hoffman, Peter Thiel, and Amazon Web Services, and Infosys of course — are supporting this great endeavor, and in addition to Alan, a great set of leaders will serve as advisors.  Ilya Sutskever, who has worked over the last several years as a leader in developing the so called “deep learning” techniques, will direct the research at OpenAI, and will be joined by several dynamic leaders and individual contributors, whose conviction and imagination will help this field move forward, based on the best of what we know, the best ideas, the best inventions, and the key lessons.

AWS is very much known for AI. Infosys is not known for anything by providing bodies, and I have been told by a person who worked for Infosys that Brightwork Research & Analysis contains more intellectual property than all of Infosys. That may or may not be an overestimate, but it is undoubtedly true that no one in the industry talks about Infosys in any degree of innovation. Infosys is an H1-B factory that draws from the least skilled of the H1-B employment pool.

Why Infosys?

Another question that’s been asked a lot, is why Infosys?  We at Infosys, with over 150k software engineers, are unique beneficiaries of and contributors to this endeavor.

No, this is false. Infosys is not known for AI in any shape or form. Vishal is doing PR, but none of it is ringing true.

The Real Infosys

Vishal Sikka presents a very “jolly” persona.

However, Infosys is a horrid company as we cover in the articles How Infosys and Tata Keep Indentured H1-Bs, and The Broad Scale Wage Theft by Tata Consulting and Infosys, and How Infosys Violated B-1 Visa Law and Charged Clients a 98.6% Margin.

Knowledge at Wharton

After Vishal left Infosys, he late popped up, having raised $50 million for an AI startup. Mindless VCs are throwing money at companies that can enunciate the phrase “artificial intelligence,” and Vishal qualifies.

He was able to get covered in the prestigious publication Knowledge at Wharton in the article How a New Startup Aims to Use AI to ‘Amplify Humanity,’ and this interview immediately reduced our impression of Knowledge at Wharton.

AI for Amplifying Humanity?

Sikka believes AI has the potential not just to transform business but also to “amplify humanity,” as he puts it. He sees AI as a force multiplier that can tackle issues ranging from climate change to self-improvement. “I would love to see tens of millions of people to be able to build intelligent systems, and billions to be able to bring basic intelligence into anything that they do,” he says.

Notice Vishal’s faux concern for “humanity” once again. This is a pattern with Vishal that he frequently demonstrates. Also, notice that “intelligence” is now classified by Vishal as artificial intelligence. Billions should be able to bring “basic intelligence” to things they do because, well, they have a brain. If we begin using the term “intelligence” to be a proxy for “artificial intelligence,” then it seems as if we have lost the plot.

Infosys is No Longer the Place for AI?

I always have been passionate about the idea of technology being a human amplifier, something that improves our ability and makes us see more, do more, and be more. Both [my wife] Vandana and I are deeply committed to this idea. In the course of last year, Vandana and I thought and talked a lot about this and concluded that the solution lies in startups. She is working on her own startup, in the consumer area. I thought that after working in large companies for 16 years, and leading them in many ways, it was time [for me] to go back to the drawing board.

Wait, as can be seen from earlier quotes, back in 2015, Vishal made it very clear that Infosys was the place for AI. And that Infosys had 150,000 developers. Surely Vishal would not change his mind from 2015.

Well, yes, it seems that after he was fired from Infosys that he has changed in mind — and “back to the drawing board” means his own $50 million startups.

Is it us, or does it seem like Vishal’s views of where the opportunity and future lies — is entirely dependent upon what company he is working for at the time? Vishal followed this same pattern previously when leading up SAP’s HANA database initiative, widely lampooning Oracle’s competitive offering, then years later when he joined the board of Oracle, switching to talking up Oracle as we cover in the article Is Oracle Board Member Vishal Sikka a World Expert on AI?.

Vishal’s Previous Employers Were Legacy, and the Future is Now Vishal’s AI Startup?

I was fortunate to participate in two transformational journeys in both those companies. Large companies have massive scale, and they have many benefits because of that scale. Every successful company does things in big ways, reaching large numbers of customers and being able to affect the work of massive numbers of people. [However], they also face significant challenges in making transformation happen. The burdens of legacy are real.

Not in the case of Infosys.

Infosys has no legacy of AI, so it is difficult to see how never having been prominent in something, this can be considered “legacy.”

Using SAP’s Definition of SAP…Against SAP?

By the way, SAP, a company that Vishal is referring to, has, for decades, used the definition that “legacy” is any software that is SAP, as we cover in the article How SAP Used and Abused the Term Legacy. Vishal referred to databases that were not HANA as legacy. Now Vishal seems to be accusing SAP of being legacy (which, by the way, is something we agree with as we cover in the article How To Best Understand SAP as Legacy for Software Companies). However, we don’t use the term inconsistently, as does Vishal.

It is curious to see Vishal call his previous employer’s legacy or having a great deal of legacy. If Vishal thought this, then why was he selling people on the idea that Infosys was the future of AI? Well, because he was working for Infosys at the time. So, when working for Infosys, Vishal trumps up the 150,000 developers (only a small portion are focused on AI BTW, which he never explained), then when pitching his startup, all of this is now legacy, and one needs to start from a blank sheet of paper. Vishal did not leave Infosys because he concluded that Infosys was “legacy,” he lost some political battle that got very public and very nasty.

More Faux Intellectualism from Vishal

David Patterson and John Hennessy, who won the 2017 Turing Award, in their Turing lecture last year talked about how it’s a fundamentally different age now of semiconductors. The traditional Moore’s Law is being replaced by domain-specific architecture for chips, especially for AI.

Here we go again. Vishal is back to trying to impress people with “big ideas.” There are now AI-focused CPUs, which we don’t see much potential in, because processing is not the limitation of using AI.

Secondly, Moore’s Law died several years ago, and this is not related to AI-focused CPUs. It is difficult to see what the relationship between these two facts is. However, Moore’s Law dies when the size of pathways on chips fell to slightly larger than the width of an electron (meaning no more miniaturization using this approach was going to occur). This led to the development of multicore processors, which face their limitations — namely heat buildup. It is a bit curious how vacuous Vishal’s statements are. They sound like something, but then when analyzed, turn out to not be any insight.

What Insight?

Vishal has had little to do but study on the topic of AI. He says he has been developing fantastic IP in AI for two years. If true, then why does he seem to be operating at a very high level of abstraction and reaching for things to try to sound insightful.

The fact is that Vishal received his Ph.D. back in the mid-1990s, and has been more involved with corporate intrigue and politics than any technical work. We noticed this same lack of familiarity with details back when he was the CTO at SAP, which is not much of a technology company and is dominated by marketing and sales, meaning that the CTO position is more of a marketing position than really a CTO position.

Rewriting History

Vishal has a curious explanation for why he is so uniquely qualified to lead an AI startup.

I have a Ph.D. in AI and have had the opportunity to work in large companies in enterprise software and services. I understand transformation in a way that few people do because I have lived through two large-scale, successful transformations. I thought it was time to take advantage of the unique gifts I have been given. That’s how I ended up here.

Yes, as I said, back in the 1990s. There are many people with PhDs in AI — are they all to be the recipients of $50 million in VC funding?

Also, to what transformation is Vishal referring to?

SAP degraded as an organization while Vishal worked there, with more employees and work being done in India, leading to terrible support, and then Vishal getting tied up in what Teradata asserts is IP theft to build HANA as we cover in the article How True is SAP’s Motion to Dismiss the Teradata Suit. (And Teradata further says in court documents that Vishal Sikka was entirely on board with the IP theft.) I have been following SAP since 1997, and I don’t know what transformation Vishal is referring to.

Vishal then left SAP unexpectantly and under unusual circumstances. Next, he took the top position at the horrible Infosys — a firm which does nothing but US worker displacement and rigs the H1-B visa program for which they were found to have defrauded as we cover in the article Who Got the $34 Million Fine from the Infosys H1-B Fraud Case?

When Vishal started working for Infosys, they were an H1-B mill that engages in H1-B fraud and is known for nothing except low priced, low skilled IT bodies. And after Vishal left, Infosys is known…..for the same thing.

What is the transformation again?

Vishal ended up at his startup because he has a name, and if you whisper “AI,” and you have a name, you can raise dumb money. How silly is a lot of the VC money? Well, let us see this quote.

“You know why VCs can’t find deals worth a s***? It is because they have a bunch of idiots and children with literally no experience doing anything in their life running their entire screening process that just chunk any business or area they don’t understand or sounds dumb without nuanced understanding of the market.” – Reddit

Vishal is making his journey sound premeditated when it was reactive.

Vishal Discovered/Realised the Weakness of AI?

This next claim is quite significant, but also quite unspecific.

I always thought that teaching and doing are great ways of learning, so I went back to school. I taught two classes last year on AI – one in California and one in China. I also worked on a few prototypes over the course of the year and realized(emphasis added) that there was both a tremendous weakness in the current state of AI, and also a tremendous opportunity to do something better.

If there is some big weakness to AI, it is guaranteed that Vishal Sikka, a person who has barely worked in AI since his Ph.D., would not be the person to find it.

And The Weakness Is….Something Everyone Who Works in AI Already Knows

You hear all these wonderful examples of things that AI techniques can do, but on the other hand, there are also remarkable weaknesses. The systems are incredibly powerful perceptual engines, but they have no idea about what is happening in the world, what is the meaning of various concepts or entities, the semantics, of how things work. The techniques have significant weaknesses.

There is no one working in AI who does not know this.

Vishal can’t “realize” or say discover something that everyone working in an area already knows about. In the same way that Vishal cannot, for instance, realize or discover New Mexico. Everyone, at least who lives in the US, knows about New Mexico (and many have heard of the great balloon festival shown above). People critique European explorers for “discovering” the New World, which was already populated with indigenous people — however, at least European explorers can be said to have discovered New World land for the people of the Old World.

Why is building context into AI such a problem?

Well in the example of humans or animal brains, they are not general-purpose computers, but instead, purpose-built processing systems.

This mouse, with its much smaller and more simple brain, is far better than you are at navigating through a maze than you are. Biologists have given up on trying to develop an animal hierarchy of intelligence because it depends upon the task being measured.

The Monarch Butterfly migrates from Mexico to Canada. They have tiny brains (the size of a sesame seed), yet they do it, even though it takes multiple generations to get there. How “smart” is this insect?

Higher-level animal brains build up an understanding over decades, and the brain is designed to be very efficient because it relies upon heuristics and does not calculate every possibility. This is easily observed by merely asking Alexa a non-standard question, and finding it does not know how to react.

As a person who has run ML algorithms for forecasting, there is a great overestimation of many of the technologies by calling them “AI.” In reality, the only work in an automated way for a short period until human intervention is required.

2001: Space Odyssey was made back in 1968 and projected capabilities out in 2001 or 19 years ago. However, 51 years later, there is still nothing even like this. Related to this application, we are now at the point where we can skip typing questions into Google, and instead ask them to Alexa or Google Home. We might recommend remaining AI into something more modest and closer to what “AI” can do in the foreseeable future. The term machine learning or ML is also similarly overly ambitious, techniques that are categorized as ML, merely output metrics that are interpreted by a human. So it seems that it might be called “computer-enabled human learning.” But then it is not sufficiently differentiated from any other use of the computer. 

Alexa could not deny us entry through the bay doors even if she wanted to. 

Putting context into AI is an enormously labor-intensive. So unless Vishal has figured out a way to accelerate this contextualization, and we are not holding our breath, then this observation will not mean much. If the author were being interviewed by Knowledge at Wharton, I would never provide this explanation, because it is simply too obvious to anyone who knows about the topic that it is not an insight.

Also, where is the interviewer to ask a follow-up question? The interviewer accepts every single answer that Vishal gives.

The response to this statement by Vishal illustrates how flaccid this interview is because this exact point should have been brought up to Vishal.

Not Enough People Know AI?

To further exacerbate that inherent limitation, you have this massively asymmetric situation where a very small number of people understands AI techniques. There are roughly 35 million or so programmers in the world, but the number of machine-learning engineers is in the few hundreds of thousands. Tencent did a study which showed that there are about 300,000 AI engineers in the world. These are people who can build machine-learning applications. The number of people who can explain to you how particular algorithms work is only about 20,000.

Why would so many AI engineers be necessary?

Recall that the idea is to place AI into software, not to have vast numbers of AI/data scientists placed at accounts billing hours (the way that IBM runs its AI “business”). Secondly, Data Science is one of the hottest fields in IT today, with hoards of Data Scientists becoming upskilled as time passes, new certifications, new degree programs, etc..

This is reminiscent of the other complaint about AI, which is that the data is not available as we cover in the article How Many AI Projects Will Fail Due to a Lack of Data? Here IBM sells AI projects, and then after they have overpromised and have already billed a number of hours, tell the client that they should have known they do not have the data to run the AI.

AI projects are not delivering anywhere close to their promises, and many are complete write-offs.

Given the enormous amount of lying about the potential of AI by all of the major consulting firms, many vendors, etc.. (which Vishal is further adding to so he can raise his $50 million), how exactly would AI projects be able to meet expectations? 

This means that in the second stage of AI, after the luster has come off of AI projects, there are now explanations for what is needed to improve this success rate. More skills. More data. More investment…etc., and fill in the blank.

However, something no AI consulting firm or Vishal Sikka will admit is that the evidence for the enormous potential and benefits of AI was never presented. It was merely asserted.

Vishal….a Man of the People?

We can do better. We can bring AI to everybody, we can build AI systems that are transparent, easy to explore, easy to understand, and explainable in the sense that the developer understands. The technique itself is inherently opaque in these deep neural networks but you can make it easy for the developer, and for the designer of the system to understand what is happening in their system by making the system exploratory and transparent. We can dramatically simplify the footprint that these AI systems carry, and we can bring problem-finding into the mix, by bringing business people, the domain experts and IT people into the construction and evolution of intelligent systems, not only the data scientists and the AI engineers.

How will Vishal make AI transparent? Algorithms are run against data, and then the values or descriptive statistics are read. One tries to find better scores and then use the algorithms that have better scores.

Is Vishal proposing some type of visualizer?

Again, there are no specifics.

From another dimension of transparency, which is openness, why is AI not transparently open? AI is mostly based upon open source and the publishing of AI algorithms. Is Vishal’s startup about creating an open-source project, or charting for his AI platform? How can a closed source entity like Vanai critique the primarily open-source AI area for lacking transparency?

You can get started with TensorFlow for free. 

AWS offers very low cost AI tools

Or one can download R Studio and run a wide number of AI/ML algorithms from within it, saving scripts as files for future use.

What precisely about AI is lacking in transparency?

Again, the Knowledge at Wharton interviewer does not ask, Vishal could say that the Moon is made of cheese, and it seems like Knowledge at Wharton is good with whatever Vishal wants to place in the article.

Furthermore, Vishal makes it sound like AI will move from a specialized area to involving everyone in the company. This sounds like its “inclusive,” but the problem is that simple statistics are not inclusive of the non-technical parts of the company. How can I be so confident in this assertion? Because for several decades, I have worked in analysis, and I have explained and presented statistical concepts to a wide variety of clients. It is challenging. This is reminiscent of the projections around visualization tools like Tableau, and that as soon as people could see more graphics, it would lead to better decision making.

Really?

Because the World Health Organization shows the US as having the most expensive health care system in the world, but having the health outcomes of a country which spends 1/5th as much and ranks 37th in the world.

This information is publicly available and has had precisely zero impact on changing the US health care system. And US politicians routinely state, without the worry of contradiction, that the US has the best health care system in the world.

Accessibility of information always improves decision making?

Vishal’s Vanai…A Brave New World of AI, an AI Platform!

I decided I wanted to build a company that on the one hand helps enterprises build AI systems that can help them become better both at growth as well as operations. On the other hand, I wanted to build that using a new platform that is built from scratch, that is empowering, accessible to tens of millions of developers, system designers, analysts and so forth, and on which enterprises can reliably and transparently build dozens of applications. We spent the last two years getting to that point.

Vishal speaks just like a politician.

Lots of high minded ideals and promises. Secondly, he is proposing that his startup will allow tens of millions of developers, system designers, and analysts to integrate. Vishal worked as CTO for SAP and then CEO of Infosys — he has no history of creating anything like what he is describing.

The following quotation from Markian Jaworksy provides some more analysis of this point.

The number of people using TensorFlow is much higher than any AI product he has released to the public.
During the Google I/O Conference in June 2016, Jeff Dean stated that 1,500 repositories on GitHub mentioned TensorFlow, of which only 5 were from Google TensorFlow – Wikipedia
https://en.m.wikipedia.org › wiki › TensorFlow

Is Vishal Platform Agnostic?

Vishal claims that his startup is platform agnostic. But the reason for this is explained by Markian Jaworsky.

“We are platform agnostic” because he doesn’t have one, and will never build one to compete with AWS or Azure.

Copying Empathy and Design Thinking from SAP?

SAP uses both the concept of empathy and design thinking to market its software. Both are nothing more than marketing constructs. SAP is a cutthroat company that is entirely lacking in empathy for its customers. The term was used by SAP to make it appear it was responding to customers after it tried to apply harsh indirect access rules on its software retroactively. Design thinking is a confusing mismatch of unproven claims around how one should design things that SAP has a difficult time even explaining, as we cover in the article Does Design Thinking Improve SAP Implementation Speed?

However, notice Vishal adopts both in his explanation of his company.

AI has the potential to transform every company, every business and every industry. The potential for adoption of AI in the enterprise represents a seismic shift, of the same magnitude as the transition to the use of computing itself. However, we believe technology alone does not create opportunity; rather, when people harness technology, they achieve more than ever before. At Vianai, we bring technical expertise and the power of empathy to our clients. We combine technical know-how with an empathetic, human-centered and iterative/experimental design thinking approach to problem selection, definition and change management. This purposeful combination of AI and Design Thinking will pave the way for innovation, productive use, and widespread adoption of a next-generation of AI-driven technologies and approaches within companies across all industries and around the world.

A big word salad of nice-sounding ideas, along with taking marketing constructs, that never proved to be anything real from SAP. Secondly, Vishal had an opportunity to show his empathy at Infosys, and he did not do it. See this scandal related to wage theft in the article The Broad Scale Wage Theft by Tata Consulting and Infosys. Is this how you show empathy to workers by stealing wages? Knowledge at Wharton could have evaluated the history of SAP and Infosys (where he was not some minor employee who could not affect policy). Still, Knowledge at Wharton was not interested in doing this.

AI Startups Are All Unique

There are several thousand AI startups…. I’m sure they’re all unique and doing fantastic things. We have assembled a team that has a unique background in that we have a deep understanding of enterprises, and of transformation in enterprises.

That is curious because, to us, this is an enormous waste of capital given the actual opportunities in AI. Secondly, there is little reason for AI startups to require so much VC involvement. Again the costs of the AI tools are not very high, and most software does not require large numbers of developers. The only companies that employ large numbers of developers are companies like SAP and Microsoft that have extremely low developer productivity.

Making Progress on Vanai?

Vishal’s company is named Vanai.

This is a page from the website. 

This is mostly marketing talk, just like the interviews given by Vishal. 

This is a firm that has raised $50 million already, and we count roughly five web pages, and along with those five pages, there are a grand total of 283 words for the entire website. After two years of study, Vishal has 283 words on his website?

At around the 26-minute mark in this video, there is a demo of the Vanai concept. 

Vishal Sikka Raised $50 Million on…..?

This is an opportunity for Vishal to showcase his company. Knowledge at Wharton asked no challenging questions, and apparently has interviewers who add no value to the proceedings. If this article were published in Forbes, we would have concluded it was a paid placement. We read a few other articles at Knowledge at Wharton, including Is Now the Best Time to Have a Baby in Corporate America, and the coverage is hugely tilted towards the status quo and to reinforcing the legitimacy of established entities. There is not much analysis that takes place at the magazine or questioning of assumptions.

However, even when faced with a compliant interview, there seems to be little reason for anyone to have given Vishal $50 million to start his company.

Vishal receives our Golden Pinocchio Award for his claims around AI. 

Conclusion

Vishal Sikka’s story around AI at Infosys and AI at Vanai does not hold up to scrutiny. This recent analysis is congruent with previous analysis of Vishal’s statements that he is an entirely unreliable source.

References

How a New Startup Aims to Use AI to ‘Amplify Humanity’

https://en.wikipedia.org/wiki/AI_accelerator

https://en.wikipedia.org/wiki/Animal_cognition

Is Now the Best Time to Have a Baby in Corporate America?

*https://www.reddit.com/r/venturecapital/comments/eri5jl/the_reason_vcs_are_just_tired/

https://vishalsikka.blogspot.com/2015/12/openai-ai-for-all.html

Vishal Sikka: Why AI Needs a Broader, More Realistic Approach