Industry Research Notes for Machine Learning

Ian Hellström

Highlights

Last updated: Monday, 21 June 2021

$1 trillion

global AI market in 2028

Grand View Research (June 2021)

87%

of AI and ML initiatives never make it to production

VentureBeat (July 2019)

55%

of organizations have not deployed a single ML model to production

Algorithmia (Dec 2019)

40% of organizations spend more than

90 days

deploying a single ML model to production

Algorithmia (Dec 2020)

95%

of production ML systems is 'glue code', not model code

Google (2014)

Cloud-native ML platforms have only been developed in the last

5 years

Databaseline (Sept 2020)

880+

cloud-native tools

CNCF (Feb 2021)

270+

AI and ML tools

Linux Foundation (Feb 2021)

50+

end-to-end ML platforms

Databaseline (Feb 2021)

Market 

Note Type Source Categories Date
70% of companies will employ hybrid or multi-cloud management technologies by 2022 prediction McKinsey edge 2021-06-15
Global AI market is expected to reach $1t in 2028 with 40.2% CAGR prediction Grand View Research 2021-06-07
The impact comes from the last 20% of the journey: 87% of leaders (vs 23% of all others) spend more than half of their analytics budgets on landing the last mile insight McKinsey 2021-04-23
Leaders spend more on proactive change management insight McKinsey 2021-04-23
Leaders invest >15% of their IT spend on AI insight McKinsey 2021-04-23
Leaders have structured data governance vs siloed POCs insight McKinsey 2021-04-23
Leaders have enterprise-wide AI talent acquisition vs pockets of AI talent insight McKinsey 2021-04-23
Leaders have a high degree of automation in data engineering/science with a modern architectures that make extensive use of unstructured data and external sources insight McKinsey 2021-04-23
Companies fall into three groups: leaders (8%), aspirants with 1.8x (17%), and laggards (75%) insight McKinsey 2021-04-23
26% have AI in production (+1 pp), 35% are evaluating AI (+ 2 pp), 26% are considering AI (-2 pp), and 13% are not using it (-2 pp) insight O'Reilly 2021-04-19
79% of leader expect to use enterprise open source for emerging technologies to increase over the next two years: 72% IoT/edge (vs 55% today) and 65% AI/ML (vs 48% today) prediction Red Hat 2021-03-03
72% expect container use to grow prediction Red Hat 2021-03-03
Barriers to enterprise open source: support (42%), compatibility (38%), security (35%), talent (35%) insight Red Hat 2021-03-03
90% use enterprise open source insight Red Hat 2021-03-03
83% of technology leaders are more likely to select a vendor who contributes to the open-source community insight Red Hat 2021-03-03
The ML platforms market is expected to grow to $126.1b by 2025 prediction Cognilytica 2021-01-20
Global AI software market expected to grow to $37b by 2025 prediction Forrester 2020-12-10
42% have a hybrid cloud/on-premises infrastructure insight Algorithmia 2020-12-09
Successful AI adopters use automated tools to produce and test AI models: 48% vs 20% insight McKinsey 2020-11-17
Successful AI adopters use advanced processes (e.g. data operations, microservices) to deploy AI: 57% vs 23% insight McKinsey 2020-11-17
Successful AI adopters use a standardized tool set to create production-ready data pipelines: 44% vs 23% insight McKinsey 2020-11-17
Successful AI adopters use a standardized end-to-end platform for AI-related data science, data engineerring, and application development: 40% vs 20% insight McKinsey 2020-11-17
Successful AI adopters use a standard protocol to build and deliver AI tools: 33% vs 16% insight McKinsey 2020-11-17
Successful AI adopters track a comprehensive set of KPIs to measure the impact of AI initiatives: 52% vs 27% insight McKinsey 2020-11-17
Successful AI adopters rapidly integrate internal structured data for use in AI initiatives: 56% vs 28% insight McKinsey 2020-11-17
Successful AI adopters have standard tool frameworks and development processes in place: 51% vs 19% insight McKinsey 2020-11-17
Successful AI adopters generate synthetic data to train AI models to complement data sets: 49% vs 16% insight McKinsey 2020-11-17
Explainability is identified as the third most relevant risk to enterprises: 41% in 2020 vs 39% in 2019 insight McKinsey 2020-11-17
Companies with more than 20% AI contributions to EBIT have: 1) better overall performance, 2) better overall leadership, and c) more resource commitment to AI insight McKinsey 2020-11-17
22% attribute more than 5% of EBIT in 2019 to AI insight McKinsey 2020-11-17
16% have taken deep learning beyond the prototyping stage insight McKinsey 2020-11-17
61% of teams need between 2 and 6 weeks to build a model prototype; 28% need more than 6 weeks for a prototype insight Spotify 2020-11-12
38% (32%) of teams need 2 – 6 (6 – 12) weeks to go from prototype to production insight Spotify 2020-11-12
Public clouds account for 36.6% as the main location of enterprise applications: 22.3% on premises, 21.6% hybrid, and 17% in private clouds insight Lightbend kubernetes 2020-10-26
75% host the majority of their applications in cloud infrastructure: public, private, or hybrid environments insight Lightbend 2020-10-26
30% are in the early stages of Kubernetes adoption as a production platform, with 19.9% still evaluating it insight Lightbend kubernetes 2020-10-26
14.8% have no plans to adopt Kubernetes as a production platform insight Lightbend kubernetes 2020-10-26
The winners of [the model performance monitoring] market will have a bottoms-up go-to-market strategy, tailored towards a developer or data scientist user persona and have an open-source component insight Two Sigma Ventures 2020-10-21
The best [model performance monitoring] products will provide immediate out-of-the-box value and be very easy to use for different personas, without requiring months of implementation time or custom work insight Two Sigma Ventures 2020-10-21
Smaller businesses have found that building the technology is harder than it looks quotation Forbes risks 2020-10-20
IBM has deprioritized its Watson technology after drawing scorn for ventures like on $62 million oncology project that made inaccurate suggestions on cancer treatments quotation Forbes risks 2020-10-20
Companies [...] overspend on technology and data scientists, without implementing changes in the business processes that could benefit from AI quotation Forbes risks, skills 2020-10-20
Amazon canned an AI recruitment tool after it showed misogynistic biases quotation Forbes risks 2020-10-20
78% of developers claim that Kubernetes add-ons cause a great deal of pain and introduce complexity insight D2iQ risks 2020-10-20
38% of developers and architects claim their work makes them feel extremely burnt out, with 51% stating that building cloud native applications makes them want to find a new job insight D2iQ risks 2020-10-20
Only when organizations add the ability to learn with AI do significant benefits become likely (73%); these organizations facilitate systematic and continuous learning between humans and machines insight BCG skills 2020-10-20
Only one in ten companies reports significant financial benefits from implementing AI insight BCG risks 2020-10-20
57% of companies have AI pilots or deployed AI solutions in 2020 insight BCG 2020-10-20
Visa earned $3b in revenue in 2019 selling aggregated data and analytics services insight Andreessen Horowitz 2020-10-15
UPS save 100m driving miles and $350-400m per year with their ORION route optimization system insight Andreessen Horowitz 2020-10-15
Netflix generate more than 80% of content views through their ML recommendation system insight Andreessen Horowitz 2020-10-15
Airbnb increased booking conversion rate by 4% thanks to ML insight Andreessen Horowitz 2020-10-15
With 11% CAGR forecast for the value-added services market between 2020 and 2030 compared to a 0.75% CAGR for current existing services, 5G is a considered a leading source of new revenue for the industry prediction MIT edge 2020-10-07
Operators aim to deliver about 80% of enterprise requests through off-the-shelf solutions, leaving just 20% in need of a customized response prediction MIT edge 2020-10-07
Executives see the enterprise and public sector digitalization programs as the most compelling market opportunities for 5G prediction MIT edge 2020-10-07
Addressing the seemingly infinite number of enterprise opportunities, with all of their hardware and software complexities, means that operators cannot go it alone prediction MIT edge 2020-10-07
The data center market is expected to grow year-over-year through 2024 prediction Gartner 2020-10-07
Spending on global data center infrastructure is projected to reach $200 billion in 2021, an increase of 6% from 2020 prediction Gartner investments 2020-10-07
Operators with serious ambitions for 5G have a cloud-first strategy for the network and IT insight MIT edge 2020-10-07
There is an ML industry shift from R&D to operations prediction Nathan Benaich and Ian Hogarth 2020-10-01
PyTorch is preferred for research, whereas TensorFlow is preferred in industry insight Nathan Benaich and Ian Hogarth 2020-10-01
ML industry shift from R&D to operations insight Nathan Benaich and Ian Hogarth 2020-10-01
AI investments have not decreased due to Covid-19 insight Gartner covid, risks 2020-10-01
There is a large group of non-tech companies that are just starting to dip their toe in earnest into the world of data science, predictive analytics and ML/AI. Some are just launching their initiatives, while others have been stuck in 'AI purgatory' for the last couple of years, as early pilots haven’t been given enough attention or resources to produce meaningful results yet quotation Matt Turck (FirstMark) 2020-09-30
Orchestration engines are seeing a lot of activity. Beyond early entrants like Airflow and Luigi, a second generation of engines has emerged, including Prefect and Dagster, as well as Kedro and Metaflow quotation Matt Turck (FirstMark) 2020-09-30
Just like 'Big Data' before it, ML/AI, at least in its current form, will disappear as a noteworthy and differentiating concept, because it will be everywhere quotation Matt Turck (FirstMark) 2020-09-30
An increasing need for real-time streaming quotation Matt Turck (FirstMark) 2020-09-30
A number of large corporations are starting to see the results of their efforts. They typically embarked years ago on a journey that started with Big Data infrastructure, and has evolved along the way to include data science and ML/AI quotation Matt Turck (FirstMark) 2020-09-30
Global spending on AI is to grow to $110b in 2024 prediction IDC investments 2020-08-25
Global spending on AI is to be $50.1b in 2020 prediction IDC investments 2020-08-25
Global spending on AI was $37.5 in 2019 insight IDC investments 2020-08-25
AI developer stacks today are like DIY craft kits, with the instructions and 70% of the parts missing quotation Clemens Mewald (Databricks) 2020-08-18
61% believe AI will substantially transform their industry in the next 3 years prediction Deloitte 2020-07-14
AI will add $13 – 16tn to the global economy by 2030 prediction PwC and McKinsey investments 2020-06-11
AI models are like six-year-olds during quarantine: They need constant attention... otherwise, something will break quotation Forrester 2020-05-28
90% of tech executives see AI/ML as key to be incorporated in products insight Bain 2020-05-26
87% of tech companies are not satisfied with their use of AI insight Bain 2020-05-26
54% growth in global AI market in 2019 insight Statista 2020-05-13
$10.1b revenue of global AI software market in 2018 insight Statista investments 2020-05-13
In 2022, AI technologies will be adopted in (compared to 2019): 1. Customer service (73%, +13 pp), 2. Sales and marketing (59%, +26 pp), 3. IT management / AIOps (50%, +3 pp), 4. R&D (44%, +4 pp), and 5. Finance (31%, +5 pp) prediction MIT adoption 2020-04-28
More than half identify as mature adopters of AI technologies insight O'Reilly 2020-03-18
26% plan to have formal data governance processes and tools by 2021 insight O'Reilly risks 2020-03-18
59% meet their ML needs by adopting enterprise software with built-in ML insight Deloitte adoption 2020-03-01
Enterprises are between early adopters (2020) and early majority (2022/2023) on the technology adoption lifecycle for cloud native technologies insight Loodse adoption 2020-01-13
75% have explored or adopted ML technologies insight IBM adoption 2020-01-03
45% with 1000+ employees and 29% with <1000 employees have adopted AI insight IBM adoption 2020-01-03
33% plan to deploy ML models in 2020 prediction H2O 2019-12-13
50% will not have deployed a single ML model in 2020 prediction Algorithmia risks 2019-12-13
20% have deployed ML models insight H2O 2019-12-13
The top ML priority for companies with more than 1,000 employees is cost reduction insight Algorithmia 2019-12-13
The top ML priority for companies with fewer than 1,000 employees is generating customer insights and intelligence insight Algorithmia 2019-12-13
Smaller companies deploy models faster than larger, if they deploy models within 30 days insight Algorithmia risks 2019-12-13
65% with 5,000 – 10,000 employees have no models in production insight Algorithmia 2019-12-13
55% have not deployed a single ML model insight Algorithmia 2019-12-13
21% are evaluating use cases insight Algorithmia 2019-12-13
17% are starting to develop models insight Algorithmia 2019-12-13
17% are ready to deploy models to production insight Algorithmia 2019-12-13
The space of ML operations ("MLOps") is just now emerging and will no doubt be big news in the next few years quotation Forbes 2019-12-12
AI specialists and ML engineers (1st: 74% annual growth), data scientists (3rd: 37% annual growth), data engineers (8th: 33% annual growth) are among the top-15 emerging jobs in the US insight LinkedIn skills 2019-12-10
30% use ML for exploration and insights generation (i.e. not production) insight Kaggle 2019-12-02
Explainability, reliability, and fairness of models are more critical in the enterprise than in consumer products, where the risk of bad predictions is minimal insight CMSWire 2019-12-02
ML adoption: 95% North America, 72% Asia, and 64% Europe insight Finances Online adoption 2019-08-27
The main arc here, which has been playing out for years but seems to be accelerating, is a three-phase transition from Hadoop to the cloud services to a hybrid/Kubernetes environment quotation Matt Turck (FirstMark) 2019-07-02
In this new multi-cloud and hybrid cloud era, the rising superstar is undoubtedly Kubernetes quotation Matt Turck (FirstMark) 2019-07-02
$28b spent globally on machine learning applications in 2019 Q1 insight Statista investments 2019-05-10
$14b spent globally on machine learning platforms in 2019 Q1 insight Statista investments 2019-05-10
Through 2022, only 20% of analytics will deliver business value prediction Gartner risks 2019-01-03
Through 2022, 85% of AI initiatives will continue to fail prediction Gartner risks 2019-01-03
Through 2020, 80% of AI will remain alchemy prediction Gartner risks 2019-01-03
Computational resources will increase 5x from 2018 to 2023 prediction Gartner 2019-01-03
By 2023, 70% of AI workloads will use containers using a serverless model prediction Gartner 2019-01-03
Machine learning platforms are of high important to traditional decision makers (e.g. CIOs) compared to medium importance to decision makers associated with a business unit (not associated with IT) insight Forrester 2017-12-06
80% of views on Netflix come from recommendations insight Mobile Syrup 2017-08-22
99.9% of spam is caught by Gmail's AI insight Wired 2015-07-09
75% of views on Netflix come from recommendations insight McKinsey 2013-01-01
35% of purchases on Amazon come from recommendations insight McKinsey 2013-01-01

Adoption 

Note Type Source Categories Date
Nothing will stifle an AI deployment like a lack of alignment on the purpose of AI itself quotation Forbes 2021-05-21
Top barriers to adoption: talent (19%), data quality (18%), use case identification (17%), company culture (14%), and infrastructure (12%) insight O'Reilly risks 2021-04-19
Only 30% of business leaders consider AI adoption to be fast insight NTT Data 2020-11-20
Covid-19 has increased adoption of AI in healthcare from 45% to 84% insight Intel covid, healthcare 2020-11-18
AI adoption in retail is 42% across the board: 33% in apparel and footwear, 29% in grocery and food, and 26% in home improvement insight Noodle AI retail 2020-11-17
70% of financial services firms use ML to predict cash flow events, fine-tune credit scores, and detect fraud insight Deloitte finance 2020-10-31
Security, privacy concerns, and the complexity of integrating AI within existing infrastructure, are at the main barriers to AI adoption insight Gartner risks 2020-10-01
The main barriers to adoption for financial services are: 1) lack of available data, 2) budget constraints, 3) lack of talent, 4) data privacy concerns, 5) lack of trust, 6) lack of available platforms and tools, 7) lack of management support, and 8) lack of algorithm transparency insight PwC finance, risks 2020-05-01
In 2022, AI technologies will be adopted in (compared to 2019): 1. Customer service (73%, +13 pp), 2. Sales and marketing (59%, +26 pp), 3. IT management / AIOps (50%, +3 pp), 4. R&D (44%, +4 pp), and 5. Finance (31%, +5 pp) prediction MIT 2020-04-28
59% meet their ML needs by adopting enterprise software with built-in ML insight Deloitte 2020-03-01
Enterprises are between early adopters (2020) and early majority (2022/2023) on the technology adoption lifecycle for cloud native technologies insight Loodse 2020-01-13
75% have explored or adopted ML technologies insight IBM 2020-01-03
45% with 1000+ employees and 29% with <1000 employees have adopted AI insight IBM 2020-01-03
ML adoption: 95% North America, 72% Asia, and 64% Europe insight Finances Online 2019-08-27
77% struggle with business adoption of AI insight NewVantage risks 2019-01-01
Business functions with significant value from AI: manufacturing (57%), risk (51%), supply-chain management (47%), product/service development (45%), strategy and corporate finance (39%), service operations (39%), sales and marketing (35%), and human resources (33%) insight McKinsey 2018-11-13
Lack of trust in AI among patients (36%) as well as clinicians (30%) is a barrier to adoption insight Intel healthcare, risks 2018-07-02
54% expect widespread adoption of AI until 2023 insight Intel healthcare 2018-07-02
Leading sectors in AI adoption: high tech and telecommunications (32%), automotive and assembly (29%), financial services (28%), energy and resources (27%), media and entertainment (22%), transportation and logistics (21%), CPGs (20%), retail (19%), healthcare (16%), education (16%), construction (15%), professional services (14%), and travel and tourism (11%) insight McKinsey 2018-04-04

Investments 

Note Type Source Categories Date
Retailers spend on average 7% of their IT budget on AI. A quarter of large retails spend up to 10%. insight Noodle AI retail 2020-11-17
74% of AI applications in retail are spent on operations and only 26% on customers-facing applications insight Noodle AI retail 2020-11-17
Spending on global data center infrastructure is projected to reach $200 billion in 2021, an increase of 6% from 2020 prediction Gartner 2020-10-07
Global spending on AI is to grow to $110b in 2024 prediction IDC 2020-08-25
Global spending on AI is to be $50.1b in 2020 prediction IDC 2020-08-25
Global spending on AI was $37.5 in 2019 insight IDC 2020-08-25
53% spent more than $20 million during the past year on AI and talent insight Deloitte skills 2020-07-14
AI will add $13 – 16tn to the global economy by 2030 prediction PwC and McKinsey 2020-06-11
$10.1b revenue of global AI software market in 2018 insight Statista 2020-05-13
US DoD funds at to the tune of $4b in 2020 insight PTC defence 2020-03-01
JAIC (Joint Artificial Intelligence Center) has increased its budget from $70m in 2018 to $209m in 2020 insight PTC defence 2020-03-01
DARPA spends $2b over the next five years to foster innovation in AI for critical DoD processes insight PTC defence 2020-03-01
$28b spent globally on machine learning applications in 2019 Q1 insight Statista 2019-05-10
$14b spent globally on machine learning platforms in 2019 Q1 insight Statista 2019-05-10

Risks 

Note Type Source Categories Date
Top barriers to adoption: talent (19%), data quality (18%), use case identification (17%), company culture (14%), and infrastructure (12%) insight O'Reilly adoption 2021-04-19
46% do not version data sets insight O'Reilly tools 2021-04-19
Traps of AI operationalization: 1) 'Ideating about AI' is more useful than actually 'implementing AI', 2) Taking a 'set it and forget it' approach to AI, 3) What works in AI pilots will scale in production, 4) Waiting to hire the perfect 'unicorn' to operationalize AI projects, and 5) The new tool we buy will help us scale our AI initiatives quotation Gartner 2020-12-09
Only 11% of organizations deploy a model within a week insight Algorithmia 2020-12-09
Only 11% of organizations can put a model into production within a week insight Algorithmia 2020-12-09
64% of organizations need a month or longer to deploy a model insight Algorithmia 2020-12-09
56% of organizations struggle with governance, security, and auditability insight Algorithmia 2020-12-09
40% or organzations need three months or longer to deploy a model insight Algorithmia 2020-12-09
38% of organizations spend more than half of their data scientists' time on deployment insight Algorithmia 2020-12-09
Smaller businesses have found that building the technology is harder than it looks quotation Forbes 2020-10-20
IBM has deprioritized its Watson technology after drawing scorn for ventures like on $62 million oncology project that made inaccurate suggestions on cancer treatments quotation Forbes 2020-10-20
Companies [...] overspend on technology and data scientists, without implementing changes in the business processes that could benefit from AI quotation Forbes skills 2020-10-20
Amazon canned an AI recruitment tool after it showed misogynistic biases quotation Forbes 2020-10-20
78% of developers claim that Kubernetes add-ons cause a great deal of pain and introduce complexity insight D2iQ 2020-10-20
38% of developers and architects claim their work makes them feel extremely burnt out, with 51% stating that building cloud native applications makes them want to find a new job insight D2iQ 2020-10-20
Only one in ten companies reports significant financial benefits from implementing AI insight BCG 2020-10-20
Security, privacy concerns, and the complexity of integrating AI within existing infrastructure, are at the main barriers to AI adoption insight Gartner adoption 2020-10-01
Only 7% say limited AI skills are a barrier of AI implementation insight Gartner skills 2020-10-01
AI investments have not decreased due to Covid-19 insight Gartner covid 2020-10-01
Frantic handoffs, manual monitoring, and loose governance impede organizations' ability to deploy more business-worthy AI use cases faster insight Forrester 2020-09-10
95% have concerns around ethical risks for AI initiatives insight Deloitte 2020-07-14
The main barriers to adoption for financial services are: 1) lack of available data, 2) budget constraints, 3) lack of talent, 4) data privacy concerns, 5) lack of trust, 6) lack of available platforms and tools, 7) lack of management support, and 8) lack of algorithm transparency insight PwC adoption, finance 2020-05-01
Industrial (safety-critical) systems are more affected by worst-case than average performance insight ML 2020-04-23
28% have a negative return on ML investments due to model deployment complexities insight IBM 2020-04-14
26% plan to have formal data governance processes and tools by 2021 insight O'Reilly 2020-03-18
50% will not have deployed a single ML model in 2020 prediction Algorithmia 2019-12-13
Smaller companies deploy models faster than larger, if they deploy models within 30 days insight Algorithmia 2019-12-13
58% have trouble scaling models up insight Algorithmia 2019-12-13
5% of companies with more than five years of experience with models in production still occassionally spend more than a year to deploy a model insight Algorithmia 2019-12-13
41% have trouble with versioning and reproducibility insight Algorithmia 2019-12-13
34% have trouble with organizational alignment and executive buy-in insight Algorithmia 2019-12-13
25% of data scientists' time is spent on deployments and infrastructure insight Algorithmia 2019-12-13
10% of companies with more than five years of experience with models in production still occassionally spend more than a 90 days to deploy a model insight Algorithmia 2019-12-13
50% spend between 8 – 90 days to deploy a single ML model insight Algorithmia 2019-12-11
87% of ML initiatives fail insight VentureBeat 2019-07-19
Through 2022, only 20% of analytics will deliver business value prediction Gartner 2019-01-03
Through 2022, 85% of AI initiatives will continue to fail prediction Gartner 2019-01-03
Through 2020, 80% of AI will remain alchemy prediction Gartner 2019-01-03
77% struggle with business adoption of AI insight NewVantage adoption 2019-01-01
Lack of trust in AI among patients (36%) as well as clinicians (30%) is a barrier to adoption insight Intel adoption, healthcare 2018-07-02
The story of enterprise Machine Learning: 'It took me 3 weeks to develop the model. It's been >11 months, and it's still not deployed.' quotation Gina Blaber (O'Reilly) 2018-03-07
74% of IoT initiatives fail insight Cisco edge 2017-05-20
When you think about data as a project, you get the Data Wheel of Death: Data Isn't Constantly Maintained → Data Becomes Irrelevant / Flawed → People Lose Trust → They Use Data Less quotation Brian Balfour (HubSpot) 2017-04-04

Skills 

Note Type Source Categories Date
Machine learning research tends to get its fair share of hype because that’s where all the cutting-edge stuff happens, all the AlphaGo and GPT-3 and what-not. But for many companies, especially early-stage ones, the bleeding-edge state-of-the-art may not be what’s needed anymore. Getting a model that’s 90% of the way there but can scale to 1000+ users is often more valuable to them. quotation Mihail Eric 2021-01-14
Companies [...] overspend on technology and data scientists, without implementing changes in the business processes that could benefit from AI quotation Forbes risks 2020-10-20
Only when organizations add the ability to learn with AI do significant benefits become likely (73%); these organizations facilitate systematic and continuous learning between humans and machines insight BCG 2020-10-20
Only 7% say limited AI skills are a barrier of AI implementation insight Gartner risks 2020-10-01
53% spent more than $20 million during the past year on AI and talent insight Deloitte investments 2020-07-14
AI specialists and ML engineers (1st: 74% annual growth), data scientists (3rd: 37% annual growth), data engineers (8th: 33% annual growth) are among the top-15 emerging jobs in the US insight LinkedIn 2019-12-10

Covid 

Note Type Source Categories Date
Covid-19 is driving investments in AI and nanotech in healthcare to grow at a rate of nearly 50% insight Vector Innovation Fund healthcare 2021-01-03
Covid-19 has increased adoption of AI in healthcare from 45% to 84% insight Intel adoption, healthcare 2020-11-18
AI investments have not decreased due to Covid-19 insight Gartner risks 2020-10-01

Edge 

Note Type Source Categories Date
By 2025, more than 50 billion IIoT devices will generate 79.4 ZB of data yearly prediction McKinsey manufacturing 2021-06-15
70% of companies will employ hybrid or multi-cloud management technologies by 2022 prediction McKinsey 2021-06-15
5G will reach 80% of the global population by 2030 prediction McKinsey 2021-06-15
56% of manufacturers will kick off edge computing pilots in the next two years insight MIT manufacturing 2021-05-24
27% of manufacturers have edge computing in production, 17% will move pilots to production in the next two years insight MIT manufacturing 2021-05-24
17% of manufacturers will move edge computing pilots to production in the next two years insight MIT manufacturing 2021-05-24
Top use cases of edge: 1) industrial IoT, 2) telco, and 3) image processing insight CNCF 2021-05-01
Top devices: 1) cameras, sensors, IoT (70%), 2) controllers and actuators (44%), 3) compute offload (25%), and 4) leaf nodes (25%) insight CNCF 2021-05-01
Top challenges: 1) security, 2) edge devices going offline, and 3) observability at the edge insight CNCF 2021-05-01
Top access methods: wifi (62%), 4G (54%), 5G (51%), CDN (21%) insight CNCF 2021-05-01
10% manage 100,000 nodes/devices or more; 49% manage <100 nodes/devices insight CNCF 2021-05-01
The power footprint for infrastructure edge deployments is to grow to 40 GW (2028) from 1 GW (2019) prediction Linux Foundation 2021-03-10
By 2028, the share of edge deployments are expected to be: 37.7% in Asia Pacific, 29% in Europe, 20.5 in North America, 7.0% in Latin America, and 5.8% in Middle East and Africa prediction Linux Foundation 2021-03-10
By 2028, 37% (down from 45%) of edge energy consumption will come from mobile (21.7%) and residential consumers (14.8%), 63% from enterprises (11.9%), telecommunications (11.9%), automotive (9.7%), healthcare (8.6%), manufacturing (6.2%), energy (4.6%), smart cities (6.1%), retail (4.5%), logistics, and transportation prediction Linux Foundation 2021-03-10
$400b to be spent on edge computing facilities between 2019 and 2028 prediction Linux Foundation 2021-03-10
Europe's 5G coverage to grow from around 1% of mobile subscriptions in 2020 to 55% in western countries and 27% in central an eastern states over the next five years prediction Ericsson 2020-12-28
China accounts for 80% of global 5G mobile subscriptions; North America accounts for only 4% insight Ericsson 2020-12-28
With 11% CAGR forecast for the value-added services market between 2020 and 2030 compared to a 0.75% CAGR for current existing services, 5G is a considered a leading source of new revenue for the industry prediction MIT 2020-10-07
Operators aim to deliver about 80% of enterprise requests through off-the-shelf solutions, leaving just 20% in need of a customized response prediction MIT 2020-10-07
Executives see the enterprise and public sector digitalization programs as the most compelling market opportunities for 5G prediction MIT 2020-10-07
Addressing the seemingly infinite number of enterprise opportunities, with all of their hardware and software complexities, means that operators cannot go it alone prediction MIT 2020-10-07
Operators with serious ambitions for 5G have a cloud-first strategy for the network and IT insight MIT 2020-10-07
By 2023 more than half of new enterprise IT infrastructure deployed will be at the edge rather than in corporate data centres prediction IDC 2020-06-01
Increased demand in government for AI at the edge insight IDC 2020-04-29
AI at the edge is 90% consumer market, not enterprise insight IoT World Today 2020-03-19
On-device intelligence is expected to be around 100% in 2025 (up from 10% in 2018): mobile devices, automotive, XR, PCs/tablets, and smart speakers prediction Qualcomm 2020-02-01
7.8b smartphone units to be shipped between 2018 – 2022 prediction Qualcomm telecommunications 2020-02-01
74% of IoT initiatives fail insight Cisco risks 2017-05-20

Ethics 

Note Type Source Categories Date
73% of tech employees agree governments ought to regulate AI insight Protocol 2021-03-15

Kubernetes 

Note Type Source Categories Date
85% of technology leaders agree Kubernetes is key to cloud-native application strategies insight Red Hat 2021-03-03
40% of end-to-end machine learning platforms on Kubernetes are based on at least one Kubeflow component insight Databaseline tools 2021-02-02
Kubeflow Pipelines (with Kale) and Katib are still in trial phase for centralized ML service, after 8 months with 2 engineers insight CERN 2020-12-01
Data scientists don't care about Kubernetes: they need abstractions that make it possible to perform high-performance data science on complex, modern infrastructure without needing to be systems expert[s] insight Determined AI 2020-11-30
It took a team of 7 people 8 months to release an alpha for Kubeflow Pipelines that integrates into Spotify’s ecosystem; beta release expected after 12 months insight Spotify 2020-11-12
Public clouds account for 36.6% as the main location of enterprise applications: 22.3% on premises, 21.6% hybrid, and 17% in private clouds insight Lightbend 2020-10-26
30% are in the early stages of Kubernetes adoption as a production platform, with 19.9% still evaluating it insight Lightbend 2020-10-26
14.8% have no plans to adopt Kubernetes as a production platform insight Lightbend 2020-10-26

Tools 

Note Type Source Categories Date
Top model/experiment trackers: custom (29%), MLflow (27%), Kubeflow (18%), Weights & Biases (8%) insight O'Reilly 2021-04-19
Top frameworks: scikit-learn, TensorFlow, PyTorch, and Keras insight O'Reilly 2021-04-19
Top deployment technologies: MLflow (22%), TFX (20%), and Kubeflow (18%) insight O'Reilly 2021-04-19
46% do not version data sets insight O'Reilly risks 2021-04-19
Python developers typically work on teams with 2 – 7 (75%) or 8 – 12 people (16) insight JetBrains 2021-03-01
Python code is run within containers (VMs) for 47% (43%) of cloud-based environments insight JetBrains 2021-03-01
Python 3 is used by 94% of developers; only 6% use Python 2 insight JetBrains 2021-03-01
Only 32% of Python developers consider themselves data scientists insight JetBrains 2021-03-01
Machine learning is the primary (secondary) activity for 49% (27%) of Python developers insight JetBrains 2021-03-01
Data analysis is the primary (secondary) activity for 48% (34%) of Python developers insight JetBrains 2021-03-01
AWS (53%), GCP (33%), Heroku (23%), and Azure (21%) are the leading cloud platforms for Python developers insight JetBrains 2021-03-01
85% of Python developers use it as their main language with JavaScript the leading secondary language insight JetBrains 2021-03-01
55% of Python developers use it for data analysis insight JetBrains 2021-03-01
54% (32%) of Python developers use virtualenv (Docker) to isolate environments insight JetBrains 2021-03-01
11% (39%) of Python developers do not use version control systems (CI tools) insight JetBrains 2021-03-01
59% of non-managed end-to-end machine learning platforms run on Kubernetes insight Databaseline 2021-02-02
40% of end-to-end machine learning platforms on Kubernetes are based on at least one Kubeflow component insight Databaseline kubernetes 2021-02-02
The most popular geospatial analysis libraries are Folium, GeoPandas, and Shapely insight Deepnote 2021-01-26
The most popular Python libraries are Matplotlib, NumPy, and Pandas insight Deepnote 2021-01-26
The most popular Python libaries for machine learning are TensorFlow (40%), Keras (34.1%), PyTorch (23.5%), and fast.ai (2.4%) insight Deepnote 2021-01-26
The most popular NLP libraries are NLTK (63%), GenSim (19.5%), and SpaCy (11.8%) insight Deepnote 2021-01-26
Python 3.6 (3.7) are used by 55% (36.5%) of users insight Deepnote 2021-01-26
Most use open-source tools and libraries for Python insight O'Reilly 2020-03-18
55% use TensorFlow, 48% use scikit-learn, and 36% use PyTorch insight O'Reilly 2020-03-18

Aerospace 

Note Type Source Categories Date
69% of leaders believe their employees are more digitally mature than their organizations insight Accenture defence 2020-09-02
The market for AI in aerospace robotics is to reach $7.78b by 2027 with CAGR 20.5% prediction Fortune 2020-08-25
Chatbots are used in 14% of aircraft and 9% of air terminals insight SITA 2019-09-07
68% of airlines want AI-powered chatbots insight SITA 2019-09-07
97% of aerospace and defence executives are willing to digitally reinvent their business and industry insight Accenture defence 2018-09-07
96% plan to increase investments to digitize core operations insight BCG 2018-05-31

Defence 

Note Type Source Categories Date
69% of leaders believe their employees are more digitally mature than their organizations insight Accenture aerospace 2020-09-02
US DoD has more than 600 active AI projects insight PTC 2020-03-01
US DoD funds at to the tune of $4b in 2020 insight PTC investments 2020-03-01
Over 30% of active DARPA projects involve AI insight PTC 2020-03-01
JAIC (Joint Artificial Intelligence Center) has increased its budget from $70m in 2018 to $209m in 2020 insight PTC investments 2020-03-01
DARPA spends $2b over the next five years to foster innovation in AI for critical DoD processes insight PTC investments 2020-03-01
97% of aerospace and defence executives are willing to digitally reinvent their business and industry insight Accenture aerospace 2018-09-07

Finance 

Note Type Source Categories Date
Top use cases in fintech and investment firms: 1) algorithmic trading, 2) fraud detection, and 3) portfolio optimization insight Nvidia 2021-03-10
Top use cases in banking: 1) fraud detection, 2) recommender systems, 3) sales and marketing optimization insight Nvidia 2021-03-10
83% believe AI is important to their company's future success with 34% citing it will increase their company's revenue by 20% or more insight Nvidia 2021-03-10
Global AI market for fintech is expected to reach $22.6b in 2025 prediction Mordor Intelligence 2020-10-31
70% of financial services firms use ML to predict cash flow events, fine-tune credit scores, and detect fraud insight Deloitte adoption 2020-10-31
The main barriers to adoption for financial services are: 1) lack of available data, 2) budget constraints, 3) lack of talent, 4) data privacy concerns, 5) lack of trust, 6) lack of available platforms and tools, 7) lack of management support, and 8) lack of algorithm transparency insight PwC adoption, risks 2020-05-01
Only 47% of finance organizations have AI in production insight KPMG 2020-02-13
Most promising AI finance applications: 1) process automation, 2) risk management assessments, and 3) fraud prevention insight KPMG 2020-02-13
Top-3 business functions that have adopted AI: 1) service operations (49%), 2) risk (40%), and 3) sales and marketing (33%) insight McKinsey 2018-11-13

Government 

Note Type Source Categories Date
NSTC's Subcommittee on Machine Learning & Artificial Intelligence must establish a task force to find best practices in identity management, such as token-based access, and single-sign-on strategies insight Fedscoop 2020-11-18
Inconsistencies in the methods of accessing and using cloud computing create barriers to using the technology for AI research insight Fedscoop 2020-11-18

Healthcare 

Note Type Source Categories Date
Healthcare must move beyond explainability to recourse (giving those impacted concrete actions they could take to change the outcome) and to move beyond transparency to contestability (allowing people to challenge an algorithm) to avoid power and participation imbalances insight Boston Review 2021-01-04
Bias is endemic in medicine, especially with regard to sex (non-male), race (non-white), class, weight, and sexuality that exacerbates automated diagnoses and decisions insight Boston Review 2021-01-04
AI and ML in healthcare are forecast to grow with 22.5% CAGR through to 2027, from a $284.38 billion market in 2019 prediction Vector Innovation Fund 2021-01-03
Covid-19 is driving investments in AI and nanotech in healthcare to grow at a rate of nearly 50% insight Vector Innovation Fund covid 2021-01-03
AI healthcare market to grow to $51.3b in 2027 with 41.4% CAGR prediction Meticulous Research 2020-12-01
Covid-19 has increased adoption of AI in healthcare from 45% to 84% insight Intel adoption, covid 2020-11-18
44% of healthcare and pharmaceuticals have increased their AI investments because of Covid-19 insight McKinsey 2020-11-17
75% are concerned AI threatens security and privacy of patient data insight KPMG 2020-02-14
More than 80% of consumers are willing to wear fitness technology insight Business Insider 2020-01-31
Most promising AI healthcare cost savings potential: 1) service operations, 2) marketing and sales, 3) risk, 4) supply chain management and manufacturing, 5) operations, 6) finance and IT, and 7) HR prediction McKinsey 2019-02-11
Most promising AI healthcare applications: 1) robot-assisted surgery, 2) virtual nursing assistants, 3) administrative workflow assistants, 4) fraud detection, 5) dosage error reduction, 6) connected machines, 7) clinical trial participant identifier, 8) preliminary diagnosis, 9) automated image diagnosis, and 10) cybersecurity prediction Accenture 2019-02-11
Annual savings of AI in healthcare are expected to reach $150b by 2026 prediction Accenture 2019-02-11
Key areas of investment for AI in healthcare: digitization, engagement, and diagnostics insight PwC 2019-02-11
Top-3 business functions that have adopted AI: 1) service operations (46%), 2) product/service development (28%), and 3) supply-chain management (21%) insight McKinsey 2018-11-13
Lack of trust in AI among patients (36%) as well as clinicians (30%) is a barrier to adoption insight Intel adoption, risks 2018-07-02
54% expect widespread adoption of AI until 2023 insight Intel adoption 2018-07-02
20% of healthcare spend is wasted insight OECD 2017-01-10

Retail 

Note Type Source Categories Date
By 2023, 95% of supply vendors in consumer goods will leverage AI prediction Noodle AI 2020-11-17
Retailers spend on average 7% of their IT budget on AI. A quarter of large retails spend up to 10%. insight Noodle AI investments 2020-11-17
AI adoption in retail is 42% across the board: 33% in apparel and footwear, 29% in grocery and food, and 26% in home improvement insight Noodle AI adoption 2020-11-17
74% of AI applications in retail are spent on operations and only 26% on customers-facing applications insight Noodle AI investments 2020-11-17
Top-3 business functions that have adopted AI: 1) sales and marketing (52%), 2) supply-chain management (38%), and 3) service operations (23%) insight McKinsey 2018-11-13

Manufacturing 

Note Type Source Categories Date
By 2025, more than 50 billion IIoT devices will generate 79.4 ZB of data yearly prediction McKinsey edge 2021-06-15
50% of today's work activities could be automated by 2025 prediction McKinsey 2021-06-15
56% of manufacturers will kick off edge computing pilots in the next two years insight MIT edge 2021-05-24
27% of manufacturers have edge computing in production, 17% will move pilots to production in the next two years insight MIT edge 2021-05-24
17% of manufacturers will move edge computing pilots to production in the next two years insight MIT edge 2021-05-24
77% of semiconductor companies have adopted or are piloting AI technologies insight Accenture 2019-08-21
Top-3 business functions that have adopted AI: 1) manufacturing (49%), 2) product/service development (39%), and 3) service operations (27%) insight McKinsey 2018-11-13

Telecommunications 

Note Type Source Categories Date
7.8b smartphone units to be shipped between 2018 – 2022 prediction Qualcomm edge 2020-02-01
Key areas of AI: network planning, network performance management, SLA management, product lifecycle management, network management, and revenue management prediction Ericsson 2019-05-16
53% of service providers expect to have integrated AI into their networks by the end of 2020 prediction Ericsson 2019-05-16
Top-3 business functions that have adopted AI: 1) service operations (75%), 2) product/service development (45%), and 3) sales and marketing (38%) insight McKinsey 2018-11-13

Utilities 

Note Type Source Categories Date
Top-3 business functions that have adopted AI: 1) service operations (46%), 2) product/service development (41%), and 3) manufacturing (19%) insight McKinsey 2018-11-13
Only 23 percent of utilities executives see AI as a high strategic priority insight Roland Berger 2018-03-13
40% lack a strategy or targets for AI insight Roland Berger 2018-03-13