JUL 2016 Machine Learning in Industry Consensus
 Share
The version of the browser you are using is no longer supported. Please upgrade to a supported browser.Dismiss

 
View only
 
 
ABCDEFGHIJKLMNOPQRSTUVWXY
1
TechEmergence JUL 2016 Machine Learning in Industry Consensus
2
Gaining ROI from Applying Machine Learning
3
Total Responses: 25 (out of 31 total participants)
4
5
Response TypeNo. Of ResponsesPercentageRank
6
Sufficient Data1646%1
7
Pick the Right Problem1029%2 (Tied)
8
Data Science Talent1029%3 (Tied)
9
10
Name (Q1) What are the criterion needed for a company to derive maximal value from the application of machine learning in a business problem? (No more than 2 sentences)Category 1Category 2 (if applicable)
11
Helgi Páll HelgasonQuantity and quality of a company's business data is vital when it comes to machine learning application. The ability to automate and integrate such applications, once developed, with business processes and workflows is also required.Sufficient Data
12
Massimiliano VersaceOwn the data, customize the algorithms to extract the best data from data, end keep improving it. Sufficient Data
13
Roman V. YampolskiyAvailability of big personal data to customize solutions to particular users.Sufficient Data
14
Slater VictoroffHaving a large amount of unstructured data, and a valuable business problem to solve.Sufficient DataPick the Right Problem
15
Dr. Steve OmohundroMachine learning works best with lots of training data. It's even better if the system can train itself by repeatedly exploring the task domain.Sufficient Data
16
Dr. Alexander D. Wissner-GrossPossession of large amounts of proprietary data relevant to their business is the key criterion for companies wanting to exploit machine learning.Sufficient Data
17
Dr. Danko NikolicThe company has to set clear goals and ensure that data exist for achieving those goals. Sufficient Data
18
Dr. Justyna Zander(1) access to proper data; (2) leverage of machine learning results used on proper data for meeting business demand that the company is addressing. Sufficient DataPick the Right Problem
19
Dr. Reza ZadehA rich dataset that other businesses don't have, since all businesses have access to the same algorithms, datasets are the differentiator.Sufficient Data
20
Prof. Anima AnandkumarAvailability of data and means to procure them efficiently: either clean data in small amounts or noisy data in large quantities. Clear quantifiable means to benchmark and test the effectiveness of the methods.Sufficient Data
21
Dr. Pieter J. MostermanWhat is needed is an abundance of data, a well defined problem, and a large set of correct answers to the problem. Additionally, ample lead time is needed to train the machine before expecting any value.Sufficient Data
22
Dr. Timur LuguevCompanies have to store not just a lot of data, but also think more about structure of stored data and best ways to clean it. Sufficient Data
23
Dr. Philippe PasquierThe main ingredients required for machine learning to be used are simple:
1- You need enough data (this is the most expensive part).
2- You need staff that are proficient with ML technics to implement it, but most importantly to have the vision and confidence to use it.
Data Science TalentSufficient Data
24
Professor James HendlerMachine-learning works best when there is clear data, preferably clean, and of large scale, that directly relates, or correlates well, with the practice of the business. Trained personnel are essential.Data Science TalentSufficient Data
25
Eyal AmirAccess and rights to use data, good peopleData Science TalentSufficient Data
26
Daniel Berleant, PhDThe organization must have a sufficient body of data (hence the catch phrase "big data") for a machine learning effort to be able to extract and learn from. In order to do this, there must be a qualified person on hand to work on applying the machine learning methods.Data Science TalentSufficient Data
27
Dr.-Ing. Aureli Soria-FrischA company should count with staff who are experts in all stages of the data analysis pipeline. This mostly includes experts not only in machine learning but also in the sensor and their measures, which will be used as features, e.g. sales forecasting needs a sales expert in the team not only data scientists.Data Science Talent
28
Dr. Patrick Ehlen, Chief Scientist, Loop AI LabsData always arrives in some context, which means a set of assumptions has been made that will bias the outcome of machine learning; The most reliable results come from analyses that are able to take more context into account as part of the learning procedure. And then management needs to be able to put aside their own biases when data-based insights reveal unexpected outcomes, such as breaking into new markets or changing their approaches to products or customers.Data Science Talent
29
Dr. Bruce MacLennanCompanies need to understand the foundations of machine learning, so they don't apply it naively and get bad results.Data Science Talent
30
Peter Boltuc, ProfessorEliminate jealous engineers from decision processes concerning deployment of creativity machines and other cognitive architectures.Data Science Talent
31
Dr.Tatsuo NakamuraMachine learning is a very powerful tool, and quite complex at times. Companies need to make sure they have the right people, with the right intentions, working with, and controlling, this tool. To derive maximal value, a company must assign very capable people to work with AI and machine learning applications with a clear objective, and make sure there are no “Mavericks” amongst them. Data Science Talent
32
Dr. Mika RautiainenGood theoretical background in data science and artificial intelligence. Knowledge how to apply theoretical frameworks into practical use.Data Science Talent
33
Prof. Amir ShapiroA company needs to find a real world application in which users can benefit from AI. Examples may be personal assistance, user intention recognition, speech and vision recognition, etc.Pick the Right Problem
34
Joscha Bach, PhDIdentify allocation problems that can scale in new ways when automated and optimized, and identify tasks (such as UI/UX, data processing in existing products) that are relevant for success of the product, and have not been touched in the last 24 months: these are very likely to benefit from incorporating recent progress. Pick the Right Problem
35
Dr Richard DowneThe most promising opportunities for machine learning applications are those that elude easy codification into rules, but which are nonetheless tedious for a human. Pick the Right Problem
36
Evion KimIt is crucial to choose which target metrics(precision/recall etc..) to optimize in machine learning model because there exists a trade-off relationship between them. Also, clearly understanding the connection between each target metric and the business goal is important.Pick the Right Problem
37
Ben WaberIs this issue rapidly changing, high value, and measurable?Pick the Right Problem
38
Andras KornaiYour ML has to be better than anybody else's ML, and the payoff will be direct in those areas which are already dominated by AI, such as high frequency trading. In all other areas, it is sufficient to be better than humans, or just as good or nearly as good but cheaper than human labor. Pick the Right Problem
39
Claude TouzetMachine learning relates to learning, i.e., discovery of relations between the data that can not be provided by humans (too expensive). Therefore, (machine) learning must only be applied to cases where the knowledge is not complete (happily there are plenty of applications).Pick the Right Problem
40
Dr. Edward ChallisThe best setting is an operational loop that is closed under automated processes. A classic example would be product recommendations in e-commerce -- a virtual environment for an AI agent to learn, through experimentation, how to up and cross sell.Pick the Right Problem
41
Dr Sean Holden(no response)n/a
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
Loading...
 
 
 
Question 1 - ROI of ML
Question 2 - ML Misconceptions
Question 3 - Industry Disruption
Question 4 - Low Hanging Fruit
All Responses
Disclaimers