Siri and Alexa have made our lives easier by taking on mundane tasks. They are phenomenal technological advancements that save your precious time, allowing you to focus on more important things in life. But have you ever wondered what the consequences are if Artificial Intelligence was misused?
Artificial intelligence is slowly but surely seeping into every domain imaginable. Technology giants world over are spending large sums in improving AI every year. Google made history with AlphaGo, the first neural network powered AI to defeat the world Go champion – Lee Sedol. AI developed by Alibaba and Microsoft have both successfully beaten humans in the Stanford reading comprehension test. But as the race continues, the moral and ethical implications of building AI are being questioned by scientists and philosophers alike.
There are many alarming statements and open letters by men of science floating around that can leave people like us grappling for answers in a complex world built on math. The Malicious Use of Artificial Intelligence : Forecasting, Prevention, and Mitigation is an excellent report penned by 26 experts that talk about the major issues and how to mitigate them while developing AI.
As a consumer of AI technology, it is imperative that you know about the challenges and consequences of adopting it in our lives. Outlined below are the 5 major issues to consider while adopting AI in our lives.
1. Misuse of AI
In the wrong hands, AI can be weaponized resulting in attacks through both physical and digital channels. It can be used to automate cyber-attacks such as phishing or exploiting speech synthesis for impersonation. It can perpetrate physical attacks through deployment of drones or some other form of autonomous weapons systems
More alarmingly, it can also be used to analyse massive amounts of data on human behaviour, beliefs and traits to create targeted propaganda. This could undermine the social fabric of democratic nations that use public debates. We have already seen the influence of firms like Cambridge Analytica that used data from Facebook users from an online poll. Imagine if there is a fairly advanced AI that can process all data being collected online. Doesn’t it make you feel quite vulnerable?
2. Algorithmic Bias
Artificial Intelligence uses algorithms like the K-NN, random forest, SVM etc. to perform the complex calculations to produce the results. Some of the algorithms are black boxes where we do not fully understand what factors influence their results yet. These algorithms often pick up human biases that are often incorporated in the code or from the data that is collected through machines or humans. This bias can infiltrate the results produced by the AI and disproportionately impact marginalized groups of the society. Therefore, it is necessary to employ proper review of the results to ensure such biases are weeded out.
3. Issue of Responsibility
In the earlier days it was easier to attribute responsibility of a hardware or software malfunction to a human or machine. But with the implementation of AI, it is no longer easy to identify perpetrators. There are usually three responsible parties when it comes to AI:
- Data scientists who develop the artificial intelligence tools
- Field experts who collect/provide the training data
- End users of the tool
It is very difficult to isolate who the responsibility falls on in the event of a malfunction. Legal proceedings will get complicated if a dispute arises.
4. Ethical Challenges
AI does not check the correctness of the data that is fed to it. It simply performs computations on the dataset without making fact-checks. This is a truly dangerous characteristic because negative traits inherent in the data like racism, sexism etc can inadvertently manifest in the results also.
Microsoft had to shut down its AI powered chatbot ‘Tay’ because it sent offensive tweets and posted reprehensible images after merely 24 hours of going online on social media. Therefore, AI results can be wrong and damaging to certain fractions of the society.
5. Data Privacy
As data processing power increases by leaps and bounds with the help of AI, the pressing problem is that of data privacy. Even the commonly available feature of facial recognition can be used by malicious groups for attacking people. Governments need to introduce legislations restricting the processing of personal information. This is quintessential to avoid discrimination of individuals based on factors like race, gender, or sexual orientation.
Summing Up
Although there are multiple challenges in adopting AI, we cannot overlook the benefits of using this technology. The challenges can be solved through careful design considerations that safeguard the interest of the users.
AI experts need to be cautious while designing AI applications to fully adhere to safety protocols and procedures to ensure that we reap the benefits alone. On the other hand, legislators need to define stringent regulations to ensure data security of individuals and organisations. This approach will surely set us on the path to using AI securely in the near future.
