The New Rules of Build vs. Buy in An AI World
As artificial intelligence as a service (AIaaS) emerges, the build vs buy debate is clear.
Since the field of AI, and specifically computer vision, has evolved to a point of being accurate and scalable enough to be useful in real-world applications, enterprise businesses have begun to realize the tremendous efficiency and revenue gains that are possible. These types of process efficiency gains are hard to ignore, and as a result, enterprise businesses are seriously evaluating the spectrum of possible AI strategies for the first time.
The problem for many businesses is that it’s difficult to make an informed decision when there are so many possible solutions. There is the traditional ‘build in-house’ route dominated by open-source toolkits like Caffe, Theano, and Tensorflow, or the full-service consulting approach led by IBM, with the artificial intelligence as a service (AIaaS) approach somewhere in-between.
This is why the engineering and product management teams that we work with at Clarifai often look to our expertise to help them make sense of the noise. When we help businesses with their AI strategy, there’s just one question that really matters:
Are you an AI company?
By that, I mean, ‘Is AI your single core competency?’ Do you have teams of data scientists and machine learning engineers dedicated to this cause alone? Are you willing to dedicate engineering time away from improving your product features, understanding your consumers’ behavior, or building core infrastructure?
If the answer if yes, then utilizing the myriad of open-source machine learning toolkits that exist and building and maintaining your AI in-house makes sense. If the answer if no, and your development teams are focused on building products and features that support your e-commerce, media, content, or social platform, you are better off choosing the best AIaaS service for your needs. Here are three reasons why:
1. Minimizing time to market to maximize the impact of your AI strategy
Building machine learning in-house takes considerable time, effort, and upfront investment. You need a team of data scientists, the right infrastructure, tons of data, and a lot of time to build a production-ready machine learning solution. And it’s not just about the time spent actually building and optimizing the machine learning solution itself - business leaders should also factor in the time involved in hiring the right people and getting the budgets approved for multiple line items.
2. Keeping up with continuous innovation and paying down technical debt
The AI space is evolving rapidly. New papers and concepts are being introduced regularly and staying on top of the research that is required for best-in-class AI can be daunting. Having a team of R&D experts at hand who are dedicated to understanding, incorporating, and introducing new technologies and concepts to an AI platform is a huge benefit of working with an AIaaS partner dedicated to building solutions in this space. Furthermore, AI is not a one-size-fits-all solution, so leveraging experts who can train and customize models to suit your unique business needs is invaluable in getting the most business value out of AI.
Another consideration for businesses building machine learning into their workflows and customer experiences is ongoing maintenance and technical debt related to machine learning. A recent Google Research paper argues that “Machine learning systems have a special capacity for incurring technical debt because they have all of the maintenance problems of traditional code plus an additional set of ML-specific issues. This debt may be difficult to detect because it exists at the system level rather than the code level.” Using AIaaS removes a layer of technical debt and helps businesses scale and future-proof their AI solution.
“Machine learning systems have a special capacity for incurring technical debt because they have all of the maintenance problems of traditional code plus an additional set of ML-specific issues. This debt may be difficult to detect because it exists at the system level rather than the code level.” - Google, 2015
3. Leveraging shared infrastructure and expertise
Both compute power and analytics tools are key to scaling any AI initiative. Working with a partner that analyzes millions of visual pieces of content per day has huge advantages. There’s no need to worry about the cost of monitoring and scaling an IT backend infrastructure as volumes grow and technologies evolve that require more compute power. There’s also the additional benefit of being tapped into a network of data that far exceeds your own, and being able to apply the aggregate learnings from this data to your own unique use case.
Any business can benefit from partnering with AIaaS, whether they have an in-house machine learning team or not. For businesses without an in-house machine learning team, AIaaS can provide a full end-to-end solution for your AI strategy by enabling your development team to take advantage of machine learning with a simple API. For businesses with in-house machine learning teams, AIaaS can allow those teams to focus on the fun, rewarding parts of machine learning without having to worry about the hassles of maintaining infrastructure.
To recap, if you’re not an AI company, don’t build it yourself. AIaaS lets you focus on your core competencies while leveraging all of the best things AI can let you do. AIaaS allows you to get all of the benefits of building machine learning in-house (and more!) and none of the hassle.
相關推薦
The New Rules of Build vs. Buy in An AI World
As artificial intelligence as a service (AIaaS) emerges, the build vs buy debate is clear. Since the field of AI, and specifically computer vision, has ev
Getting the pixel coordinates of text or ticks in matplotlib
otl nat gen enum reference title info bsp http The exact pixel coordinates of title, labels, legends or ticks are important information f
Xenoceratops, A New Species of Horned Dinosaur Discovered in Ca computer nada
www.inhandnetworks.com Scientists in Canada have named a newly found horned dinosaur (ceratopsian) and it’s called Xenoceratops foremostensis. I
What the new Mastercard logo will look like in China
What the new Mastercard logo will look like in ChinaImage from ‘Brand New’ blogIn Brand New blog, the new Mastercard is described as:the logo is easily ide
The New Wave of Digital Transformation | AITopics
aitopics.org uses cookies to deliver the best possible experience. By continuing to use this site, you consent to the use of cookies. Learn more » I und
AI Is The New Face Of Systemic (And Automated) Inequality
One source of income inequality is prejudice. Unconscious (and, sadly, conscious) attitudes direct opportunities more to favored groups and steer them away
The new geopolitics of artificial intelligence
The multilateral system urgently needs to help build a new social contract to ensure that technological innovation, in particular artificial intelligence (
Mastering The New Generation of Gradient Boosting
Mastering The New Generation of Gradient BoostingCatboostGradient Boosted Decision Trees and Random Forest are my favorite ML models for tabular heterogene
How to Reins GRUB After the New Update of WINDOWS
smo 文件 自己 重建 orm date ans 輸入 div la ji windows OK 現在我的輸入法也裝好了,再說一遍,垃圾windows 這次大更新又把我的GRUB搞崩了 這次沒有像以前一樣用chroot重建grub馬上搞定,所以記一下網上找到的方案: 1.
【理解】column must appear in the GROUP BY clause or be used in an aggregate function
pin group func 函數 ear -- database nbsp like column "ms.xxx_time" must appear in the GROUP BY clause or be used in an aggregate function -
MIT is investing $1 billion in an AI college
Ever since the beginning of the AI boom in the early 2010s, there's been a corresponding drought in talented AI developers and researchers. The way to fix
BigGAN: A New State of the Art in Image Synthesis
“Best GAN samples ever yet? Very impressive ICLR submission! BigGAN improves Inception Scores by >100.”The above Tweet is from renowned Google DeepMind
【MongoDB】The basic operation of Index in MongoDB
drop desc ould lar text and tracking num ack In the past four blogs, we attached importance to the index, including description and c
ES6: for...of VS for...in
highlight 範圍 ons dom erro 原型 設計 tor 數字 for...of和for...in是js中常用的兩個遍歷的方法,但是它們遍歷是有區別的,最主要的區別就是: (1)for...of是遍歷key, value中的value即鍵值,for...of一
The Usage of Lambda and Heap in the C++ STL
ner class eap cto con c++ stl nts been nta The Usage of Lambda and Heap in the C++ STL Heap In c++ STL, the heap had been implemented as
Qualcomm platform, the commonly used parameters of charger and battery in device tree file
inpu res alc max stc sum In nat TE 1 battery charging voltage : qcom,float-voltage-mv = <0x10fe>; 2 battery recharge threshold : qc
UserWarning: The installed version of numexpr 2.4.3 is not supported in pandas and will be not be used
color true mini nbsp 4.6 minimum img sed ack 問題:/usr/local/lib/python3.5/dist-packages/pandas/core/computation/__init__.py:18: UserWarni
The current state of generics in Delphi( 轉載)
var ont ask logic ted win else ddd calling The current state of generics in Delphi To avoid duplication of generated code, the compil
Controlling Industrial IoT Connectivity the Spin Polarization of Electrons in Three Dimensions
www.inhandnetworks.com The interior bulk of a topological insulator is indeed an insulator, but electrons (spheres) move swiftly on the surface
資料視覺化之"Surveying the complementary role of automatic data analysis and visualization in knowledge discovery"
Surveying the complementary role of automatic data analysis and visualization in knowledge discovery Enrico Bertini; Denis Lalanne; VAKD '09 Proceedings o