What’s real and what’s hype in Artificial Intelligence and Machine Learning
As with all new technology, emotions run wild and our excitement gets the better of us. The hype is everywhere — projections of an AI-ruled planet akin to something out of a science fiction movie and computers that can almost read minds. In the midst of all these, entrepreneurs would like to know exactly what AI can do for their business.
Now that the initial excitement is over, and the dust has settled a bit, it is easier for us to have more level-headed discussions about Artificial Intelligence and Machine Learning. This article attempts to explain what’s real and what is still wishful thinking as far as these two new technologies are concerned.
One advantage of our rapid adoption of technology globally is the vast quantity of data being generated. But this has also created a problem – too much data.
Machine learning, a vibrant research area under artificial intelligence (AI), helps us make sense of this data. Machine learning is a breakthrough in programming which enables the computer look for patterns in data and attempt to learn from them.
By learning, the computer becomes smarter in the process and is able to sort more patterns effectively. This effectiveness is then applied to real life scenarios.
Here are the major things machine learning can currently do:
Computers are now able to recognise objects, faces and places in pictures and video, and even label them with an impressive level of accuracy.
Sequential data and speech recognition, translation
Computers can now recognise speech, acoustics, handwriting strokes, and even translate from and into various languages, made possible by something called recurrent neural networks.
Machine learning enables computers analyse medical records and activity, e.g. prediction of disease progression, for therapy planning and support, and for overall patient management. It is also being used for early disease detection.
Computers can now place information in classes which enables analysts to make better decisions. This is particularly useful in the financial services sector. For example, before a financial institution hands out a loan, it evaluates customers on their ability to repay the loan by considering different factors such as earning, age, savings and financial history.
Information Extraction (IE)
Machine learning enables the extraction of structured information from unstructured data sources such as web pages, articles, blogs, business reports, and emails. Already, companies like Work Fusion provide financial institutions with AI powered solutions which automate the digitization and categorization of unstructured documents (like invoices, bills, declarations, certificates, letters), extracting and matching relevant data which saves the company thousands of hours of manual labor.
These “abilities”are not 100 percent flawless and are still undergoing improvement. However, we are at a stage where these abilities are adequate enough to be useful in the real world.
Some of these real world applications include:
A combination of speech and face recognition enables intuitive security systems to be built. Security systems can also register and recognize handwritings, signatures, fingerprints, iris/retina identification and verification.
Digitally enabled businesses are able to make more accurate cross selling recommendations and offers to their customers based on the vast amount of data being crunched by the computers everytime the customer interacts with the business.
Responding to service inquiries in voice and chat conversations, finding mistakes in financial records audits, sorting through huge amounts of data and structuring them.
Computers aid medical practitioners with disease discovery and patient monitoring.
Machine learning algorithms enable farm robots to identify weeds and apply herbicides with pinpoint precision.
Self driving cars consolidate a lot of the abilities machine learning has afforded us in one place – image and sound recognition, robotics etc. With successful test drives in different countries, self driving cars are projected to be broadly deployed from 2020 onward.
. . .
Other areas include elements that you already experience in your daily life e.g. improved spell-checks and autocorrect when you’re typing an email or SMS on a smartphone, phone assistants like Siri and Google Now, and iPhone’s face recognition. These are pretty exciting times indeed.
However, machine learning is nowhere near as powerful or impactful as people seem to believe. For one, it cannot run the marketing end of even a small business from end to end.
It is best to think of machine learning as a co-pilot and not an autopilot.
A human being is still needed to make judgment calls on the computer’s output. For instance, using something called cognitive automation, a machine learning-enabled software can do the heavy lifting of say, sorting through years of customer data within a few minutes. The human being plays the role of quality assurance, ensuring the machine has done the right thing. Does the pattern spotted by the computer make sense? Only a human being can make that judgement call.
We’re still at the stage where machine learning needs a lot of hand holding. In fact, right now, a significant part of the work regarding machine learning is still happening in the lab. Only the success stories make it out into the mainstream many of which have been highlighted above.
The good news is, it is an area with a lot of interest and is thus, seeing a lot of brilliant minds tweaking and improving the technology.
At the end of the day, machine learning and AI is really about the tools helping us to live better lives and be more productive. How they go about that is still relatively limited but we will keep seeing more improvements in the years to come.
Till then, we’ll need the experts to keep doing what they are doing. Just don’t expect JARVIS anytime soon.