Andrew Ng To Kickstart A New Generation Of AI

“Andrew Ng will soon be launching a campaign; a competition to push for data-centric models.”

Lots of people joke about how 80 p.c of machine studying is just knowledge cleansing. Additionally, many individuals have a look at machine studying as a glorified, technical model of statistics—a area that locations an excessive amount of significance on knowledge. If something, this tells us one factor for certain: Data is essential. Even a well-known face within the ML group, Andrew Ng has stressed how ML must take a extra data-centric stance fairly than a model-centric one.

Become A Chartered Data Scientist™
Achieve the best distinction within the knowledge science career
REGISTER NOW

Nearly 90 p.c of ML fashions constructed globally are by no means dropped at gentle, primarily as a result of they can’t regulate to the number of info out there in real-world purposes. In a 2020 survey, solely 22 p.c of corporations had made use of their fashions, lots of which took so long as 12 months to carry to customers. Traditional software program is backed by code whereas each code and knowledge allow AI programs. However, many software program builders nonetheless work on codes and mannequin architectures fairly than knowledge after they discover their ML fashions in a little bit of a repair. 

Earlier this yr, Andrew Ng brought attention to MLOps, which offers with utilising machine studying fashions in manufacturing programs. Andrew Ng believes that specializing in knowledge right here, as a substitute of solely engaged on bettering one’s code, may unlock multitudes of recent multimillion-dollar purposes of synthetic intelligence. He claims that present architectures are extremely developed for figuring out pictures, recognising speech or producing textual content. Tinkering with their structure is probably not the perfect methodology to allow them to carry out higher anymore. 

Ushering subsequent gen AI

The answer Andrew Ng has proposed is to place apart the structure of an AI mannequin and deal with what it’s working with, i.e. the information. By paying shut consideration to what a mannequin learns and bettering the standard of information, and subsequently retraining the ML mannequin, engineers can construct larger high quality programs in a a lot shorter time.

Andrew Ng will likely be launching a marketing campaign to elucidate this viewpoint on June 17th 2021. The marketing campaign will jump-start with Landing AI’s (an organization based by Ng to extend using AI in conventional industries) competitors—which can comprise contestants competing to achieve the perfect efficiency by amending knowledge in an in any other case fastened mannequin. The competitors will finish on September 4th—which simply so occurs to coincide with John McCarthy’s birthday (he got here up with the time period synthetic intelligence)—the place the highest three winners will likely be invited to a non-public roundtable occasion with Andrew Ng, himself, and have alternatives to debate their concepts and ideas with everybody current. 

See Also


Andrew Ng says that he hopes the competitors will change the many years of model-centric custom held by builders. Despite this model-centric strategy, loads of analysis backs Ng’s data-centric viewpoint. A Cambridge study reported that probably the most essential however usually neglected facet in ML fashions is knowledge dispersion. Smaller datasets must cope with noisier knowledge, whereas bigger ones make it tougher to label them. This makes for vital bottlenecks when deploying ML options into the true world. 

Keeping this in thoughts, Ng says that the shift to data-driven practices will assist remedy numerous challenges that AI at the moment faces, together with studying the best way to carry out a process from tens of hundreds of information factors (as a substitute of the present tens of millions!), studying to grasp when people don’t agree (e.g. when completely different medical specialists don’t conform to a prognosis), selecting up inconsistency amongst knowledge sources, modifications in knowledge over time attributable to one thing like modifications in behaviour, and creating helpful artificial knowledge when precise knowledge will not be abundantly out there.

Bringing this huge paradigm shift in how AI is constructed won’t be straightforward. Andrew Ng feels that it’ll require as a lot analysis and growth because the shift from ‘old fashioned AI to deep learning’ has within the current many years. Andrew Ng’s DeepLearning.AI, is initiating a course to show this data-centric strategy on easy-to-reach platforms like Coursera (curiously, additionally based by Andrew Ng). He has additionally given numerous displays on DeepLearning.AI’s YouTube channel and Amazon Web Service’s Machine Learning Summit. Andrew Ng believes that the best folks can put this concept to make use of constructively to counter many points, resembling manufacturing, treating illnesses, vitality consumption and meals manufacturing, all with the assistance of AI-backed with the suitable knowledge. 


Join Our Telegram Group. Be a part of a fascinating on-line group. Join Here.

Subscribe to our Newsletter

Get the most recent updates and related presents by sharing your e mail.

Mita Chaturvedi

Mita Chaturvedi

I’m an economics undergrad who loves ingesting espresso and writing about know-how and finance. I prefer to play the ukulele and watch outdated films when I’m free.

LEAVE A REPLY

Please enter your comment!
Please enter your name here