New Delhi, June 25: While Artificial Intelligence is not new to the world and the term was first used in 1956, an “explosion” in AI has taken place today due to increased speed of transactions and the assimilation of various technologies for networking, Pravin Anand, Managing Partner of Anand & Anand IPR law firm said.
Speaking at a webinar on ‘Intersection of AI, Copyright and COVID’ organised by LegitQuest, the seasoned IP lawyer said the World Intellectual Property Organisation (WIPO) has come up with a database called Patentscope which has 20 million technology disclosures which are relevant for Covid-19.
They have also come up with a WIPO policy tracker which lists out country-wise the intellectual property offices and what relaxations they have done in the light of the pandemic. “For example, do they allow working? Have they extended deadlines for the performance of certain tasks? Have they suspended filing of fees? All this has been laid down in the WIPO policy including the Indian Patent Office,” he said.
Mr Anand said that the WIPO has said with regard to Covid-19 that Intellectual Property does not pose a problem of access, but the policy challenge is because of the problem of absence of treatments and cures and of vaccines.
“It is only when innovation has taken place that access becomes (possible). The sequence has to be innovation followed by access. There is a need to incentivise so that the solutions can come. Only when the solutions come the access becomes an important issue. This is what WIPO has emphasised in its April 2020 papers,” he said.
Mr Anand emphasised that works generated by artificial intelligence deserve copyright protection for several reasons. “One of the reasons is that the entire intellectual property system is an incentive-based system and that incentive cannot be lost,” he said.
“Secondly, if you see section 23 of the Copyright Act, anonymous and pseudonymous works are also protected, and third, because human effort is involved behind artificial intelligence, just as much as human effort is involved in a digital camera or a smart software.”
He pointed out some core issues associated with the use of Artificial Intelligence — biases, safety and reliability, and accountability.
“Bias develops when because of copyright reasons the learning algorithms are unable to have access to certain data. So assume that there is a facial recognition AI program and the images of children are not available and only the images of adults have been fed to come up with a model which the learning algorithm creates. Obviously, in the runtime phase when questions are posed to this model the answers will not be based upon the faces of children because it will not recognise the faces of children. So, a bias will develop and this bias has a lot to do with what data is available to teach the learning algorithm to create the appropriate model,” he said.
Coming to the issue of safety and reliability pertaining to AI, Mr Anand said, “I will give an example of a case brought fourth by Microsoft. They spoke about it extensively in their books — hospitals in the US were finding correlations between asthma and pneumonia and saying that because asthma patients were the least in mortality due to pneumonia, therefore asthma patients had their belief that in conclusion they would not have admitted them to hospitals and that would have led to disasters.
“But humans went deeper and interpreted this data and their interpretation led them to believe that this was because asthma patients were getting timely care and because of timely care there was low rate of mortality, not because there was a correlation between asthma patients and mortality.”
He said safety and reliability means that humans have to work in sync in interpreting the data carefully. “Accountability, of course, is a very important issue. Those who develop these AI systems must be responsible or accountable for the results,” he said.
Giving examples of use of AI in various programmes and applications right from the Dragon Dictation software that a lot of lawyers have used in the past to drones, autonomous cars, robotics to fight fire, Blockchain, 3-D printing, alpha go, he said, “…one very interesting example of how the Mars rover curiosity selected sites for picking up samples of soil on rocks. Or how BenevolentAI is a program which is being used in drug discovery where the figures are that they go through every scientific paper which is available in print and draw Eureka moments hundred times a day and this is leading to an analysis of the clinical trial data to see why a particular molecule failed in a clinical trial and what could be done to come up with variations.”