The Best Machine Learning Company of 2021
Many impressive breakthroughs. Who comes out on top?
2021 was a great year for Machine Learning. We had a lot of developments with multiple tops and turns. The sheer number and quality of the multiple papers and outcomes released in the ML space were amazing. We had innovations in GPU, newer models, lots of research into different fields, and some ground-breaking discoveries. The Machine Learning industry continued to grow by leaps and bounds.
Here are some interesting stats. The expected value of the global machine learning market by 2027 is going to be 117.19 billion according to Fortune Business Insights 2020. According to the report, “Artificial Intelligence (AI) Hardware Market Size 2021 Analysis, Growth, Vendors, Drivers, Challenges With Forecast To 2027” just the AI Hardware Market will be USD 87.68 billion by 2027. Therefore, it should not be surprising that there are a lot of stakeholders and organizations vying for supremacy in the Machine Learning space. To prepare for my upcoming video on the State of the ML in 2021, I dug through advancements by multiple players to get a good feel for the macro of the field. I will share which organization has had the best 2021 in the context of Machine Learning.
Our players/Evaluation criteria
There were way too many revolutionary things happening with AI this year. Despite shifting a lot of the attention to their MetaVerse, FBs Machine Learning had a productive year. Google continued its prolific publishing spree, with papers from Google AI, Google Brain, and DeepMind pushing boundaries in many areas. DeepMind, in particular, had some fantastic papers, including Exploring the beauty of pure mathematics in novel ways combining ML with Math Research. The results were interesting.
On the Hardware Side, Nvidia really stepped up the hardware performance. Qualcomm and other players also had great years. This year also saw the rise of many Auto-ML startups. As such, it is impossible to have one objective answer for the best. Depending on what you prioritize, your answer will be different. I’ll be basing my analysis on how the company has set itself to take advantage of the various fields in the future. In other words, which company is best positioned to be able to take advantage of advances in the many aspects related to ML and the various stages in ML pipelines.
The Company+ Reasons
Getting right into it, the company is the tech giant Microsoft. I’ll spend the rest of the article on how MS has positioned itself to take advantage of advances in various fields. The key to understanding this is the concept of Verticle Integration. Verticle Integration occurs when a company takes charge of multiple steps in its operation. The image below is pretty good for understanding the concept.
In the case of ML pipelines, think of having your hands in multiple steps from the pipelines, from Hardware Manufacture and Data Collection to the final Deployment and Monitoring infrastructure. Working with multiple stages in the pipeline is very important for Machine Learning. To learn how to design your projects to improve your ML skills exponentially, check out this video. Let’s see how MS is involved with various stages in ML/Tech pipelines.
Stage 0: Hardware
MS is by no means a leader on the hardware side. However, their investment into Surface and other hardware components is a pretty clear indication of their intention to change that. It would not be surprising to see this change over the next years. This is their weakest aspect as of writing.
Stage 1: Data Sources
When it comes to the number of sources that MS has for their data, they’re second to none. MS owns Windows, the Worlds Most Popular OS. They also have their own Linux distro and are heavily involved in Open Source.
On the professional side, they own Github (more on this later) and LinkedIn. These are two platforms that give them access to large amounts of very valuable data. Half the organizations I’ve worked with use the MS office suite, giving them another very valuable source of data. The amount of data MS has coming in from different platforms is something that should not be overlooked. This diversity of Data Sources allows them to pivot into whatever kind of project has the highest ROI at that moment, giving them tons of flexibility.
Take a look at Github copilot. This is something that would be impossible without having large amounts of high-quality data. Their ownership of Github meant, they had access to that. Thus they were able to collaborate with OpenAI to create a revolutionary product that has gotten rave reviews and will be used by many people.
Stage 2: ML Models and Research
This is what most people think of when they think of Machine Learning. And Microsoft has had an exceptional year in this domain. Among many others, Microsoft published their study, “Fake It Till You Make It Face analysis in the wild using synthetic data alone.” They were able to write a facial features detection protocol that was trained purely on synthetic faces (faces generated by networks, not belonging to humans). Their performance on Real-human faces was state of the art.
Their performance using synthetic faces is very important. Think of cases such as facial deformities where we might not have enough sample data in the real world. Traditional Models would fail in these cases. This is why we use synthetic data to augment the input and allow our models to perform well over such cases. Using synthetic data allows for some interesting applications that I will write about in another article. For this article, it suffices to say that this paper was impressive.
MS even made waves in the field of Natural Language Processing. In their publication, “Efficiently and effectively scaling up language model pretraining for best language representation model on GLUE and SuperGLUE” they presented their newest model that beat everyone else on the NLP benchmarks GLUE and SuperGLUE. To learn more about this, check out this article.
Notably, T-NLRv5 first achieved human parity on MNLI and RTE on the GLUE benchmark, the last two GLUE tasks which human parity had not yet met. In addition, T-NLRv5 is more efficient than recent pretraining models, achieving comparable effectiveness with 50% fewer parameters and pretraining computing costs
Safe to say, MS is at the forefront of the Machine Learning Research and Implementation side of things. Don’t forget the impressive Github Copilot AI.
Stage 3: Model Deployment and Maintenance
Once a model is trained and ready to be used, it’s often shipped to the cloud. This is where MS is revolutionary. They invested in Azure and Cloud Computing in 2010. Since then they’ve continued to expand their cloud presence.
Look at the offering that they offer. By offering their high-level analytics (such as the NLP architecture in the last section) they are able to attract clients in various fields. These clients then either use the analytics as-is or customize upon them. Both options give MS lots of valuable information that lets them customize their offerings further. This makes it better, attracting more clients, causing a positive feedback loop.
Having their own cloud-based services also allows them to be somewhat protected from external failures. That’s a minor advantage that further adds to their system.
MS being so ahead in multiple fields gives them a fantastic position. Machine Learning is a complex field that relies on multiple different fields. This is even more true when it comes to actual implementation. That is why we see so many niches in the field. By being spread across so many aspects of ML, Microsoft is in a position to take advantage of advancements in ML and related fields. This is why MS is the player best positioned to come out on top at the end of 2021.
Notice that a lot of this is the cumulation of years of investment and effort. Windows is older than I am, Github was acquired in 2018, Azure was first developed in 2010, and their ML research is built from years of collaboration and learning. Rome is not built in a day, and the reason MS is in its position is due to decades of build-up. The headlines often make it seem like things happen overnight, but most often it takes 20 years of work to become an overnight success.
If you liked this article, check out my other content. I post regularly on Medium, YouTube, Twitter, and Substack (all linked below). I focus on Artificial Intelligence, Machine Learning, Technology, and Software Development. If you’re preparing for coding interviews check out: Coding Interviews Made Simple, my free weekly newsletter.
For one-time support of my work following are my Venmo and Paypal. Any amount is appreciated and helps a lot:
Venmo: https://account.venmo.com/u/FNU-Devansh
Paypal: paypal.me/ISeeThings
Reach out to me
If that article got you interested in reaching out to me, then this section is for you. You can reach out to me on any of the platforms, or check out any of my other content. If you’d like to discuss tutoring, text me on LinkedIn, IG, or Twitter. If you’d like to support my work, use my free Robinhood referral link. We both get a free stock, and there is no risk to you. So not using it is just losing free money.
Check out my other articles on Medium. : https://rb.gy/zn1aiu
My YouTube: https://rb.gy/88iwdd
Reach out to me on LinkedIn. Let’s connect: https://rb.gy/m5ok2y
My Instagram: https://rb.gy/gmvuy9
My Twitter: https://twitter.com/Machine01776819
If you’re preparing for coding/technical interviews: https://codinginterviewsmadesimple.substack.com/
Get a free stock on Robinhood: https://join.robinhood.com/fnud75
The Best Machine Learning Company of 2021 was originally published in Geek Culture on Medium, where people are continuing the conversation by highlighting and responding to this story.