Its a perfect time to start your AI business, why and how ?

AI is all the rage these days and for probably the right reasons.

In this post I will go over why do I think the timing is good for starting an AI business and what one needs to do to get started, Hint Machine learning expertise should not block you.

Why the timing is right ?

I do not cover macro economic status here, so if 2008 is to repeat soon then probably timing is not that good, but since its hard to predict the next crash timing lets ignore that for the rest of the post.

With the right timing the most critical success factor for every new startup(business wannabe) is market (the pool of customers), A good market will make your business fly, allow you to strengthen your team and grow your business naturally as it grows by itself.

How does the opportunity look like ? Machines can now see, hear and understand what was considered as humans only territory not so long ago — So what can you do with infinite amount of eyes, ears and brains ? A world of possibilities

Taking it to the next level , Automation of existing work, traditionally done by humans have a lot of upside potential both in cost saving and revenue increase. In simple words look around you, There is something someone is doing that machines can replace(They are likely to be losing their job while you creating yours if you succeed).

In dataloop.ai we see these cases every day, cases where human expertise can be replaced by machine expertise. Its all around you, Just learn the fundamentals and you will start seeing it yourself as well.

The two critical pieces of deep learning

How does AI (deep-learning from now on) get created ? The little secret of the AI industry is that it doesn't know how to creat artificial intelegence. Yes you heard me right, We dont know how to create AI, How ever we do know how to transfer intelligence from humans into machine, the wire in which this transfers happen is data.

The basic principle behind every AI out their is humans (with some expertise) are embedding small pieces of their intelligence into data pieces, the machine can consume these pieces and mimic the humans = intelligence.

So the big question you should be asking your self now is what type of expertise and data do I have access to ?. What ever it is, If you have access to unique data and expertise you are positioned to create a new, valuable technology.

The secret sauce for creating deep-learning application is therefore access to data and access to human intelligence (domain expertise) that knows how to tag that data and transfer the intelligence to machines, Lets take few examples:

• You have access to farms and the farmer knows how to identify sick cow, flower, fruit or corp.
• You have access to medical data and doctors who can analyze this data.
• You can collect text of politic posts and voting surveys results.
• You can collect images of clothes and have good understanding in fashion.
• You work on the city hall and know what is important to public health

So it all comes down to identifying market need around your expertise , collecting data and annotate(Mark) this data according to your AI business goal.

This is the operation most AI companies spending their time on and also their biggest source of intellectual property, so any data and expertise around you can be converted to AI IP, assuming humans can do the operation on the data as well to get the business value.

But wait, What about the machine learning expert ?

The industry is moving fast and the community is growing faster, Don’t shoot for complex models and you'll be fine overcoming this hurdle, When it comes to vision I usually recommend : Start with something a human (even if its the world champion) can see in 1 second. you'll be surprised how much the human brain can do in one second.

“For AI furniture company I will take a carpenter owning millions of photos over top AI talent as a partner”

Expect model sharing

The learning algorithms and frameworks is commoditizing rapidly, this means a solid single tech guy as a partner (or business if you are the tech guy) should be enough for to get you started and prove value generation, as you move on the field you have selected will need more expertise and complexity but if you are solving real world problem with market value customers will follow and good chances investors as well, allowing you to grow the team and adjust as you move forward.

Expect model training automation

All the big guys out-there are working very hard to allow you automatic generation of deep-learning , In dataloop we work with Google and Google AutoML works surprisingly well, I will expect it will get even better as time goes by, reducing the pressure for companies to hire data scientists and machine learning experts.

So look around you, AI will take jobs but it will also create jobs — can it create your next job ?