Example:

• An AI system is trained using 10 years of police arrest data to help improve efficiency.
• However, the data is biased because officers patrol certain areas more often, assuming they have higher crime rates.
• This leads to more arrests in those areas, not necessarily because more crime happens there, but because police are present more often.
• As a result, the AI learns that these areas have higher crime rates, reinforcing the original bias.

Example (referring to the previous scenario):

• Because the AI was trained on biased data, it believes some areas have more crime than others.
• It sends more police officers to patrol these areas.
• This leads to more arrests there, reinforcing the AI’s belief.
• Meanwhile, crime in other areas may be missed, which further strengthens the bias.

Environmental impact from production

• AI systems rely on large numbers of GPUs (graphics cards) to process data.
• Producing these components requires large amounts of water.
• This water is taken from the environment, which can harm local ecosystems.
• Manufacturing these components also releases CO2, contributing to global warming.

• AI systems use large amounts of electricity during their operation.
• Much of this electricity is still generated using fossil fuels.
• This leads to increased greenhouse gas emissions and contributes to global warming.
• Data centres that run AI systems also require significant cooling, often using large amounts of water taken from the local environment, which can impact ecosystems.

Deep-fakes:

• AI generated representations of real people are often referred to as "deep-fakes".
• Deep-fakes are problematic as they can portray the individual in a negative light, or even be used to profit from somebody's likeness.
• Deep-fakes raise many questions around who owns their own identity.

Misinformation:

• Misinformation is spread for either financial or political gain.
• AI tools make it much easier to produce misinformation.
• AI is now capable of generating video content that can fool the viewer into believing it.
• This poses a risk as it is increasingly easy to influence others with falsehoods.

Example:

• A supermarket uses AI-powered self-checkout systems.
• The AI scans items, recognises products, and calculates totals automatically.
• This reduces the need for staff to manually process every transaction.
• Customers are served faster, improving overall efficiency.

Example (referring to the previous scenario):

• Because the AI handles scanning and payment, staff are freed up.
• Employees can help customers on the shop floor instead.
• This improves customer service while keeping checkout times short.
• The business becomes more productive overall.

Example:

• A hospital uses AI to analyse thousands of lung scans.
• The AI identifies early signs of diseases such as cancer.
• It can detect patterns that are difficult for humans to spot.
• This allows doctors to diagnose patients earlier and begin treatment sooner.

Example (referring to the previous scenario):

• The AI analyses patient history, lifestyle, and scan results.
• It suggests a treatment plan tailored to that individual.
• Doctors can use this information to choose the most effective treatment.
• This improves recovery rates and patient care.

Example:

• A bank uses AI to analyse millions of transactions in real time.
• The AI quickly identifies unusual spending patterns.
• This helps detect potential fraud much faster than humans.
• Customers can be alerted immediately to suspicious activity.

Example (referring to the previous scenario):

• The AI learns what "normal" spending looks like for each customer.
• It flags transactions that do not match this pattern.
• The bank can block or investigate these transactions quickly.
• This leads to safer and more reliable financial services.