Subscribe to DSC Newsletter

#Enterprise #AI insights from the AI Europe (@AI_Europe) event in London

Enterprise AI insights from the AI Europe event in London

 

I attended AI Europe last week. It was a 2 day event dedicated to AI only - and it was well worth attending!  In this post, I capture my key insights from this event. I have an interest in Enterprise AI due to my course Implementing Enterprise AI

Over the two days, there were some key general themes in the event

  • AI is the new industrial revolution
  • Augmented intelligence is the near term future instead of Artificial intelligence
  • Superhuman capabilities (from machines) in niche application areas
  • Can humans handle exponential capabilities? Intellectually humans not equipped to deal with exponentials

 

Below are more specific insights including the speakers’ session names

Uber: Why Deep Learning?

For AI capabilities, we need

  • Knowledge representation
  • Ability to query inference across knowledge bases
  • Machine learning
  • Ability to understand Context through industry semantics

In the talk (Opening Speech Bringing machine learning to every corner of your business DR. Luming Wang, UBER ) had a slide called Why Deep Learning which gave a concise need for Deep Learning

 

Why Deep Learning?

  • Better results than traditional ML or human in
    • Image recognition/classification/object detection/OCR
    • Voice recognition/classification
    • NLP/Translation
    • Super long sequence pattern recognition
    • Require few feature engineering
    • More data = better results
    • More data + deeper network = even better results
    • Easy to get reasonably good results

Uber: What are the challenges of Deep learning

What are the key challenges for Deep Learning? Another Uber slide

  • There are many frameworks and each has its pros and cons (Caffe, Tensorflow, CNTK, MXNet, Theano….)
  • If you get lucky, you’ll be very happy but otherwise, you may get frustrated.
    • Hard to understand
    • Hard to debug/fix/tune
    • Few theory guidance – super high dimensionality, vast amount of parameters
    • Very long training time – often in days or even weeks
    • Expensive hardware
    • Often require customized hardware for prediction

UBS: Potential applications of AI in financial services

Financial services will be the key drivers for AI. Annika Schroder(UBS AG) had an excellent slide on use cases clustered by financial services activities (Banking: why UBS is interested in AI and other fintech innovations ANNIKA SCHRÖDER, UBS AG)

 

  • Customer Service and Engagement: Conversational Interfaces, Virtual Advisors, Customer Service, Marketing, Smart Spending, Passion Fields, Client Segmentation, Sales
  • Investment and Trading: Investment Strategies, Investment Sentiment Analysis, Investment Reporting, Quantitative Trading, Investment Research, Investment Risk Management, Knowledge Platform, A.I. Trading
  • (Cyber) Risk and Security: Cyber Incident Investigation, Intrusion Prevention or Detection, Payment Fraud Detection, Authentication, Source Code Scanning, Data Loss Protection, Surveillance,  Forensics
  • Regulatory and Compliance (RegTech): AML, Compliance Advisory, Rogue Trading Prevention, Automated Compliance Monitoring, KYC, Contract Due Diligence, Information Governance
  • Operations: Recruiting, Spend Analysis, Autonomous Documentation, Credit Risk, Automated Reporting, Invoice Processing, Vendor Management, IT Support and Infrastructure Management
  • Others: Employee Services, Expertise Network, Streamlined Mailing, Core Banking, Automated Scheduling

Cognitive Scale: Unbundling  is driving AI

Practical AI: bringing scalable machine intelligence and continuous learning to the enterprise ROBERT GOLLADAY, COGNITIVESCALE presented that everything is unbundled and that is the main driver for AI.

Nvidia: Computer vision inside

Cutting-edge research teams, hyper-scale data centers, enterprises using AI…: deploying deep learning everywhere SERGE PALARIC, NVIDIA

Serge had a few excellent slides which covered a range of themes. For me, the significant ones are

a)       The Nvidia platform scale from the Jetson (which we use) to DGX1

b)      Edge processing for the Jetson . I discussed the theme of deploying a model to the Edge from the cloud in the article Continuous improvement IoT AI learning

c)       The significance of Computer Vision in many areas including Drones

UBS: AI v.s Blockchain

Annika Schroder from UBS made a very interesting comment – for Banks,  #blockchain is collaborative v.s. #AI is competitive – I never thought of it in that way. Interesting insight

UBS: AI capabilities stack

Also from UBS – an idea of the AI capabilities stack

Uber’s AI  problem is likely to be replicated in many other industries

Uber had an interesting slide around unpredictable short term demand and a diverse range of suppliers (as the main driver for AI) . I think many more businesses would soon be in a similar situation and would need to deploy Enterprise AI

 

Typical Business Challenges

  • Lower prices
  • Faster delivery
  • Higher customer service expectations
  • Demand volatility
  • High number of products
  • Supply complexities
  • More frequent shipments

 

Uber head of Deep learning believes in Strong ai

Uber’s Deep learning head believes in Strong AI in this lifetime! (machine's intellectual capability is functionally = to human's.)  A strong statement indeed.

 

PWC virtual consultant collaborate double entry book keeping

“Deploying machine learning MICHAEL RENDELL, PWC” made three interesting comments

a)      PWC is developing a virtual consultant who already ‘works i.e. the person can advice similar to an existing human PWC consultant

b)      An AI was able to ‘learn’ double entry book keeping over a weekend after being fed some million invoices(I do not recollect the exact number)

c)      In the near future, we will have meetings with (say) 3 humans and an AI. This will be a real collaboration i.e. the AI will not be expected to just perform some tasks but people will be expected to actively collaborate with the AI

Offshoring: How many people for an outsourced/offshore project?

This is a personal observation. There are a lot of people who simply do not see AI for it’s disruptive potential. I had a conversation from a representative of a large IT offshoring company – who asked me ‘How many people would it take to typically develop an AI system?’  Contrast with PWC above! I think ICT outsourcing/offshoring companies will be the ones earliest hit by AI if they do not adapt!

Supernatural interfaces (snips) privacy by design

Privacy by design: a visionary way of using technology and data that respects confidentiality YANN LECHELLE, SNIPS  Provided an example of Privacy by design being central to future services. examples include: Connected car: “Let’s refuel on the way at the usual gas station” Location and booking: “Find me the cheapest seafood restaurant near my hotel”  etc

 

On a personal note

It was great to see long term friend from the mobile industry David Wood @dw2 . Also looking forward to reading @cccalum book t- also recommended by David The Economic Singularity: Artificial intelligence and the death of ...

Conclusion     

In conclusion, this was a great event and I very much hope it will be held in London again next year. The event reflects the rise of AI. There were many interesting insights also for my course Implementing Enterprise AI.  

 

Views: 1065

Comment

You need to be a member of Data Science Central to add comments!

Join Data Science Central

Follow Us

Videos

  • Add Videos
  • View All

Resources

© 2017   Data Science Central   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service