[Music] so far i've only talked about the positive sides of ai but ai developments have also been a source of controversy let's look at some of the issues raised one of the greatest fears is that ai could be used in unethical ways to identify and discriminate against people on the basis of race age or religion or that it could gauge our emotions to let advertisers know what we're in the mood for most major ai developers including huawei have adopted or are working on ethical principles to ensure that their technologies aren't used to cause harm even
so when an ai is sorting through data its algorithms can sometimes unintentionally learn discriminate on the basis of gender erase actually this can happen even when race and gender are removed from the raw data because ai can learn to discriminate using proxies much like humans do this can lead to unfair outcomes if an ai is used to support decision making to prevent algorithmic bias ai developers can periodically audit for biases in the decisions that their algorithms make they can also review their data sets to make sure that it's not skewed machine learning data that's causing
the ai to acquire biases another controversy with ai is the ability of the technology to potentially identify and track people who hadn't agreed to disclose their personal information ai enabled products like smartphones or smart cars for example are equipped with numerous sensors and cameras that are designed to collect data even when the information is collated anonymously an ai fed with enough data can sometimes de-anonymize users it could possibly figure out personal details like gender age religion or tastes that users hadn't agreed to disclose ict companies like huawei are taking measures to respect user privacy one
approach is federated learning in federated learning ai algorithms are trained by using data sourced from multiple decentralized network edge devices this approach contrasts with centralized machine learning where all data sets are uploaded to a main training server by limiting the quantity of data put together federated learning reduces the likelihood of de-anonymization [Music] seeing is believing right or is it a type of ai called generative adversarial network or g-a-n enables machines to perform better at image analytics or image translation in medicine it helps to detect tumors or it can convert satellite photos into maps but gan
is also being used to create extremely realistic deep fakes video sharing platforms like youtube already have several videos of famous figures like obama or tom cruise saying or doing things that they never actually said or did deep fakes can be very entertaining but obviously they can also be misused for ill intent another concern is that ai might take over people's jobs it's an understandable concern but most ai experts believe that the fear is overblown humans have always been looking for ways to be more productive the invention of the wheelbarrow the steam engine or the laptop
computer didn't result in massive unemployment ai will primarily replace jobs that are repetitive boring and don't involve complex human interactions and it will boost human productivity where it's sorely needed in areas like healthcare or quality control ai isn't able to take the place of humans where creativity or emotional intelligence are required writers software engineers and managers are a long way from being replaced by ai this said and i'm sure some of you already know this ai has successfully been used in fiction writing gptn a series of natural language processing packages can churn out new paragraphs
in a novel based on a few sentences setting the tone writers who have used the technology see it more as a compliment to their work than a threat but it gives rise to a new set of issues for example should the software company that created the ai share the authorship rights as you can see from the controversies i just described the challenges in future ai developments aren't always technological they're also legal and ethical and we also need to accept as societies that ai in the future will perform some jobs that are now done by humans
in many deployment scenarios it's already happening that's it for this part in our next and last episode we'll look at the future prospects for ai development [Music] you
