This article is by Featured Blogger Tarry Singh from his Blog Page. Republished with the author’s permission.
I was talking with a global consulting company leader the other day and he told me that they had close to a thousand “Top AI Experts”.
I was like : “Wow !”
It baffled me since we’re in the midst of the craziest deep learning projects in multiple industry verticals and we know that it's not easy to get that talent for that specific business or research project.
So I thought of doing a bit of reality check with my audience on Linkedin.
One of my contacts, who is a robotics and compute vision engineer, further shared this in his network that led to a lot of interesting discusion about : Real AI VS Hype!
Quick Intro —So What’s Going On?
Artificial intelligence is undeniably hot today, so it is not surprising that companies are throwing all kinds of jargon that matches anything and everything with AI in it.
There are a few startups who are currently building products in their narrow niches and know what it takes to build a solution from scratch.
And on the other hand there are others who are just converting their current offerings into AI projects.
You need to be able to spot the difference between the two.
Let’s first try to understand why its happening.
40% of AI Companies are Lying about their AI - Research Report
Honestly, I can bet you that that number is definitely higher that that!
AI or Artificial Intelligence is a magnet for capturing VCs and investor. It also helps companies spread the perception that they are “doing AI”.
According to this state-of-AI report published a year ago, nearly 40% of companies in Europe that claim to have AI are simply not telling you the truth .
According to the report these claims are not always made by these companies themselves but are misreported by other 3rd parties. This is not a problem but the fact that these companies do not rectify this misreporting creates the perception that they are indeed AI companies which is a problem.
The report also found that most use cases these companies describe as their awesome AI projects are in essence “quite banal”. For instance 26% of these companies have some form of chatbots, another whopping 21% use fraud detection.
While the above examples may have some basic form of machine learning deployed and might offer some relief to their clients, they are definitely not AI (read: deep learning) solutions.
And to make matters worse and clients more confusing, you have those robots, “AI powered” intelligence-powered toothbrushes?
You can download the full report here.
Fortunately media is reporting fake AI companies regularly
Thankfully media is reporting when lies go through the roof.
For instance this company reported below by the Verge. They raised $30M with SoftBank but eventually had human engineers doing all the “automated AI work!
The Guardian also ran a very interesting article investigating how minimum wage workers are put in the loop to pretend to do automated work in the backend.
As this entrepreneur aptly put it way back in 2016:
Still many companies can also go unnoticed and you must investigate further how to spot the difference between Real AI vs Fake AI.
How do you do that?
Spotting A Fake AI Company
Although it is not really very difficult, you will need some form of understanding of machine learning and deep leaning before you can start to investigate further.
But still can start investigating by asking just two questions:
1. Skills - Dig into their talent pool
I work regularly with world’s leading researchers and engineers. I hire them personally and ask point blank questions about their project experience, software development and most importantly motivation - their willingness to learn.
I dig deeper into computer vision and NLP technical skills, knowledge and experience. Then I ask questions about their projects experience. Sometimes these discussions go long and can be quite exciting.
Bottom line: If the talent pool is poor, you know enough.
2. What is their product or service?
I normally start asking questions point blank.
Are you are an AI company?
Is your product or service built with AI (read: ML or DL)?
What %age of your product uses active data (primary, secondary and tertiary data streams) (Read: what is the penetration)
How is your platform learning. (Read: Is it really AI or some software script doing update ETL batches?)
Show me please.
Then I sit back and listen.
Becoming a Real AI Company
Ok, so we’ve seen a bit how messy the world of AI startups, scale-ups and established (Read: white labeled software projects) businesses looks like.
It is also pretty clear that enterprises, startups and scale-ups are looking for a justifiable and monetarily defensible business. It is also obvious from many VCs and investors notes that the real defensible (Read: monopolistic) business model is still elusive to most startups and scale-ups.
AI companies appear, increasingly, to combine elements of both software and services with gross margins, scaling, and defensibility that may represent a new class of business entirely.
While the observations and advice of authors from a16z blog is definitely well informed, it fails to mention actionable to-do’s for competition-agnostic and sustainable data-driven business models that companies outside Silicon Valley should adopt.
Companies such as Netflix, Uber, Google, Facebook, Apple, Microsoft, Amazon and a suite of growing companies that have tasted the flavour of data, are further accelerating their growth and increasingly drawing huge margins.
How do they do this?
So, let's briefly explore what hygiene factors are needed for a company to scale from present to a data-driven company. This is based on my real-world experience out of the trenches.
Winning Strategies in an AI Economy
Step 1: Align & Drive Your Strategy and Vision
CxOs, business owners, investors and shareholders definitely need to get on board to take active interest, build awareness and take ownership AI transition or transformation.
Second, they need to sit actively and monitor its progress. CDO or CAIO “Chief AI Officer” come and go but VCs, investors and owners are burdened with the change.
Third, Ensure that your pursuit and your questions inside your organization are led with your management layer learning and educating itself to answer your hard questions with data.
Ascertain the progress by setting measurable goals so your managers and executives not only answer those questions but can demonstrate their ability to how monetizable MVPs (minimum viable propositions) that are born from data.
Bottom line: Your strategy is three words. Data. Data. Data.
Step 2: Prepare Your Staff To Become Real Machine Learning Experts, not “AI Experts”
Once strategy and vision is aligned with actionable and measurable results, it is time to educate your staff for a compounding effect so both business and technology talk the same language.
Whether you’re selling to B2B or B2C -your decisions should not be based anymore on guesses or gut instincts but data.
Managers, sales, marketing and internal teams are constantly chasing metrics that are satisfying the need of the management or even some KPIs that don't apply anymore in the AI Economy.
Corporations like Netflix, Booking.com, Uber are not running on gut instincts, they aren’t even driving guesstimates - they all are attempting to hypothesis and answer these questions with data.
Similar approach needs to be applied to help train your existing engineering, developers and technical staff in all forms of machine learning techniques. This is an essential part of your AI transformation journey.
MOOCs have been a great boon for beginners to familiarise with the concepts but that is where it stops.
Question: Ask yourself the following question: Do you think Google, Facebook, Amazon are sending their employees to MOOC platforms OR are they giving them real world data from their own pipeline?
No amount of online trainings can be substitute for our own programs in which you are applying your own insights based on your own data!
After you have prepared your organization on all possible skills that are required to scale your business with insights, it is time to put it to test!
Step 3: Don't Waste Time — Build Your First Algorithm and MVP within weeks!
Most companies spend months, if not years, in large-scale data projects to develop a data strategy, data governance, data platforms and god knows what not.
This is a typical ploy for most companies that are heavily entrenched in competitive market ecosystem and consultants are constantly pitching narrow-solutions.
As this proceeds, applications and platforms are introduced, sold and implemented internally. This makes way for consultants and skilled workers, who can only master the tools and platforms, to further penetrate these companies.
Question: Do you think this is how data-driven companies operate? Do you think Tesla is letting external platforms drive their agenda? Or do you think they are building their platforms and controlling their own destiny?
There are often data scientist inside these organizations but there is no data-centric platform or hub that has a following banner hanging: “We answer all your questions with data”.
Change this as soon as possible by building your first AI product you can pitch inside your company!
Step 4: Build Defensible business with Unique Multi-Model &-Algorithm Strategy
OK, so we’ve got our strategy and vision aligned, we got our skills update with the latest and greatest what Machine Learning and Deep Learning has to offer. We even built our first exciting algorithm and now what?
I have seen many enterprises that quickly slip into the easy chair after having reached a milestone in step 3. This commendable job but data-ambitious organizations need to realise that this is the beginning of an exciting run.
Also, it is quite easy for your competitors to catch, depending on how generic you're algorithm is.
It is time to now build a defensible business with a multi-algorithm, model-driven strategy that will involve further accelerate development of sophisticated AI engineering pipelines that do end-to-end data-intensive computations to further create market lead in their individual businesses.
Question: Companies like Uber have taken the task of data engineering extremely seriously. In the process they have ended up developing and releasing tools and platforms for doing multi-GPU , parallel processing of heavy computation loads that would be impossible for engineers to do on their own.
Step 5: Prepare to own HW, SW and Services Stack to Maintain Lead
As you continue to build your automated solutions, improving your margins and satisfying your customers’ needs, it is time also to envision where you will drive your growth based on scale (which will remain a single-digit growth margin) and where you will drive based on value (this could be your double-digit growth vehicles that could span out into new ventures).
Apple has been extremely successful with their HW+SW+Services play. Tesla is the latest example of a company that prefers to own the ecosystem of ecosystems to maintain giant lead from the beginning.
AI Economy - driven by Machine Learning and Deep Learning applications and services ecosystem, stands to disrupt and perhaps even displace AI-poor companies once and for all.
Yes, I agree that AI playbook is still being written, and that is why you need to write your own with great urgency.
While all companies try their best to avoid black holes, the key is not to steer through the competitive battle zone but aim to create a new vector of growth so as to shoot away from the competitive battle zone.
That path is that of home runs (high margin play) and industrialisation of services at scale.
It can be yours.
So let's stop faking it, and go ahead and do it for real!