SmartCat is a data company that works with other companies to help them build AI solutions. We are a blend of data scientists and data engineers and that makes us question from different angles how the next big AI module will be integrated in your platform. This is the prime reason why we can brag that we have more than dozen AI solutions in production developed over the last 8 years. In this talk we will share our experience working with various clients on AI solutions. We will give you a list of the 10 most common AI pitfalls that prevent AI solutions from ending in production. Are you familiar with PoC drawers, where RnD departments allocate some money to try something new, build that up to a working solution, but it never ends up in production? This list will help you prepare for your next AI project and it will help you lower down the chance for it to end up in the PoC drawer.
3. I AM OPTIMIST
I'm here because I've always believed in the power
of data to change the world. Being the CEO of a
service-oriented company is not easy job, but when
you have a team of young, brilliant minds pushing
the boundaries of knowledge and technology, it
gives me hope that we can turn our vision into
reality!
at the and of the day :)
47. Resources
● Newsletters:
○ The Batch - company from Andrew NG
○ FutureLoop - newsletter about innovations and AI
○ Ben’s Bites - AI trends and innovations in 5 minutes read
● Courses
○ AI for everyone - Andrew NG, AI course for non tech people
● Podcasts
○ Lex Fridman - interviews with successful people, often he calls tech and
AI champions
○ All In Podcast - investors podcast, they cover tech space
Editor's Notes
Optimist vs sceptics - important to explain you are always working with two groups and it is important to manage expectations
Ja sam Nenad Bozic, osnivac i CEO kompanije SmartCat. Na kraju dana ja sam optimista, vec 10 godina verujem da odlucivanje spram podataka, analitika i poznavanje brojki doprinosi kvalitetu vodjenja biznis i ubrzavanju/optimizaciji
Dealing with new tech has 2 problem always: expectations and process or lack of it
Story about expectations
Started in .com era and then everybody was building new and shiny websites so SEO marketing emerged as a profesion
Do you know what was the most hated question by all SEO marketing specialists
Bring me to the top of the Google search page.
Here’s some things that are easier to do than that.
VISIT THE NORTH POLE
BALANCE A TIGHTROPE
PET A WOLF
Education - importance of educating your client what is possible and what is not possible.
Similar story with AI lot of futuristic movies out there and fancy breakthroughs lot of variables, external factors, lack of education
So how to avoid most hated requests in AI
Treat this as a journey and add Safety signs in relationship in each step of the process (but do not scare them). Build trust
Good indicator that expectations are problem, even Andrew Ng created that kind of the course
The other thing with new tech, there is no firm way of working with it
Let us take web based calendar application
Your users told you they want to schedule meetings using the app, invite colleagues, and so on.
Gather the team and explain the process
Now you just need to call your SEO expert to bring it to the first place on Google right?
But what if you decide to work with tech that is climbing on popularity curve, be an early adopter. There is opportunity but there is price to pay, not knowing the best process for new things.
No expectations, no process…. No problem ----> I would argue that there is a good measurement of your success
Questions: on how many AI projects have you worked that ended up in production? Who worked on at least 1? 2-5? 5+
I will share how we:
manage expectations - with ourselves and our client
Process - work through a process that has the highest chance of success (or tells us on time to give up)
We at SmartCat have learned this the hard way, more than dozen of AI projects in production over past 4 years
We have learned our process the hard way. Over time we began to discover what factors strongly influence whether the project will be a success or a failure - how we and our clients should approach our project.
Or, to put it another way, both our clients and ourselves made lots of mistakes, and failed a lot. Thankfully, we’ve learned enough that we now have a very clearly defined process that either takes you to success, or tells you early enough that you shouldn’t waste more money on this or any AI projects.
The Pyramid, great tool to position company on the level of maturity working with data
Business due diligence + tech due diligence
Outcome - lots of lessons, a clear process and a realistic perspective on AI,
And this is a brief overview. You want to start with business understanding and for this we organize workshop to sit down with you, speak about your day at work, ideas, pain points. Next thing is to analyze the data you collect, what can we see in it, how can we use it. When we have business understanding and data understanding we can move on to the next phase, we can present you with a couple of projects you can do to improve your productivity with data you are collecting. We choose the best one and proceed with implementation, first we do web demo so you can verify we are on the right track and we then move to integration and deployment phase to bring this AI solution in production.
Process does not hide dragons sleeping along the way on each step of the way
Process looks easy, well no! Lot of potential pitfalls along the way
Easy right? Well no. Along the way there are a lot of dragons sleeping. We have learned this the hard way so I am here now to share some of our experiences and hints on how to avoid the problems on this exciting journey.
Before we begin - One important thing on how to approach your first AI project… and in case you’ve already done them, please pretend your next one is the first one :)
Example Optimus Power:
There is our product, Optimus Power, it is AI based controlling software for HVAC. We started developing it because it is cool and we wanted to play with tech, we wanted to do something for environment (CO2 emission) and in the end save money to owners of buildings.
If we asked ourselves why we should solve this problem couple of times in the beginning, we would understand that this is hard problem to solve, that we do not understand building structure, we do not have building automation experience and would probably give up or assemble a better team at the beginning.
Can you make a change with 70% ----> everybody wants 90% because they read it on the internet ----> explain commercials
For one company that is helping students apply for scholarship we have built recommender engine. Dynamic payment, we defined two results, more submits but also less seeing and then not submiting
Ubaciti foru: So, we’ve done the business analysis, the project is good to go, now we just take the client’s data and voila! Right? OF COURSE the client has data, right? I mean they’re doing an AI project, data is like the only thing you need?
.
Knowledge transfer chatbot application ----> no data ----> instead of working with them on AI we have created survey to start collecting data
First thing first, when I was talking to Gaga I realized labeled data is normal for tech people but not for the rest of the world, let us first explain that. Google Captcha
Machine like a baby, learning from images, cats and dogs
BITCOIN example, radili smo trader info portal 2017/2018 i tada je prilicno volotile bilo kretanje cena, trenirali smo prediktore na tome i odjednom stabilna prica
(napraviti kao zagontke) So, someone wanted to make an AI judge in US, and what do you think happened when it started processing cases? Was it fair, unbiased? Hint: they took the data between 1950-1990s.
Data + Business = finished product, right, well no, this is our biggest improvement over our process, first two were obvious
Post delivery phase - you need maintenance, you need constant improvement, you are building software for humans and humans change their behavior so you need to adopt. Since we do not have Artificial General Intelligence yet you need constant improvement through maintenance
Algo tuning phase relate with SEO optimization, explain new content which must be optimized, explain new customers coming to website with different habits, explain google changing algorithm
Human in the loop - AI and IA, we are not in the age of AGI yet, we are using IA (intelligent assistant) which is helping humans make a decision, use SIRI
Complementary field is Big Data, lot of processing, lot of storage, external tools - Running costs - use it when you analyze your investment, this will run on infrastructure with huge amounts of data, infra and maintenance is not cheap
Data is everywhere, it is so easy to dive into it and end up being buried with technical solution but remember that we are here building data solutions for people.
We need to think about business goals, user experience, expectation management. We need to talk a lot and understand the real need behind solution we are creating. We need a good process which will put stakeholders in every step of the process so we can tune, steer or give up in a timely manner. Remember this is an experiment, hypothesis, RnD after all and there is a chance it might not succeed. Pay attention to those mentioned pitfalls along the way. Share yours if you have them, send them to us, write a blog post, educate our future AI audience.
If you approach AI this way, you will have the highest success, you will improve the fastest and thus position yourself best in the future/reap the most rewards from AI.
Or you will conclude you are not ready yet, save your investment and put it where it’s needed more.