.png%3Ftable%3Dblock%26id%3D0aa615e0-d749-4da0-a318-aad7373b953a%26cache%3Dv2&w=3840&q=75)
Hi, thanks for visiting my post. To help you walk through the article and still relax, I first want to give you an umbrella (a very big one!), in order not to slip over a flood stream of wordy explanations for these loosely related terms.
Since this post is written in a relaxing manner - on a sweet friday night, the use of some words is just to prove some general point, but not to be exact. Hope I didn’t make you stressful so far (lmao)..let’s get started 💩
AI Engineers are expensive
In 2022, AI is less hyped than before but still, it’s a buzz word. ML Engineer is nowadays paid higher than any other type of engineers. To add the term “AI” on the resume is to add some stack of money on your monthly pay slip.
AI Engineers' salary are going to the moon, so far it can only see through telescopes.
These numbers are just the income of a typical ML employee. The full-loaded cost - the total cost that a company pays to hire and maintains that employee, is much higher. If you can hire one with low price (and he/she’s good), then you’re lucky.
AI Research are hard and time consuming
Imagine you are Aladin and step across a magic lamp. You were granted 3 wishes and the usage is as following:
- “I want to build a working (concept) product asap”
- “I want my core business not to be about AI”
- “I want to build a big AI team that delivers state-of-the-art accuracy”
The genie will slap at your face and quit his job as a wishes giver.
If you want to kickstart product fast, and your core business is not AI (say, you’re building a shopping site), then you’d better off not hiring an AI Researcher. They won’t help you get the product up and running in 1-2 month time, given that they’re also expensive.
Cloud-first ML - An alternative
Given that AI Engineers are so useless in most startups, what are the other options that you can opt for to have some AI capabilities up and running in your system?
In this case, you can leverage some pre-built AI solutions that are offered by some third party providers. For example:
- Image Recognition: Google Cloud Vision API, Azure Computer Vision
- Chatbot: Botfuel, Dialogflow on Google Cloud, IBM Watson Assistant
- Search: Google Cloud Search, IBM Watson Discovery
- Recommendation system: Cloud Recommendation Engine, BoostCommerce
Disclaimer: I’m not paid to put some brand names here (Though I wish I were)
These are the platforms that offer ML solutions with ready-to-use capability. Some fine-tuning/training maybe required but no ML expertise is required (a.k.a no-code ML), all you need is just to input some data and labels and you’re good to go.
The performance of these models on the other hand is from acceptable to very good, and that’s all you need to create a PoC product.
In the nutshell
Using pre-built ML solutions can help you start your project fast and lean, the results at the time are good enough, no attachment no contract to any AI employee what-so-ever. With this you can have your concept demo-ed and at the same time very little resource is being used.