Thursday 27 May 2021

Ethical AI: What are the Ethics in Artificial Intelligence

 

Ethical AI

When people think of advanced artificial intelligence, many think of Hollywood films such as Terminator films, Ex Machina, and Transcendence. Films, where machines imbued with artificial intelligence, have risen against humans. Such concepts seem more fiction than fact but humanity is hurtling towards a world where machines could control the world in myriad ways. Fortunately, industry watchdogs are beginning to re-evaluate the dangers of self-learning artificial intelligence (AI) encouraged by an awareness of the need for ethics in the development and deployment of AI. This brings us to the topic of ethical AI.

What is ethical AI?

In a nutshell, ethical AI attempts to tackle the ethical concerns that people have with how AI and robots are being used and designed. Today, the field of ethical AI has broadened to include theories of the consciousness and rights of artificial intelligence.


Roboethics

Robot ethics or roboethics is the concept of designing robots with artificial intelligence using codes of conduct to ensure that the automated system will respond to circumstances and situations ethically. At the crux of it, roboethics calls for ensuring that the automated system has the capacity to make its own decisions when it comes in contact with human beings. The AI should never lead to decisions or circumstances where human life, safety, and dignity are put in any danger. Roboethics is primarily concerned with the actions of the robot but it is also concerned with the thought and actions of the human developer who created the robot and AI.

Why is there a need for ethical AI?

For too long we have convinced ourselves that technology is neutral. Sadly, the reality is different. In the past, technology has been used to conduct psychological experiments on social media users by manipulating their emotions and using the data for advertising or political purposes. Here are a few more reasons why ethical AI is so important and how it can and has been misused:

  1. Biased AI:
    Biased artificial intelligence can reinforce discrimination and put minorities, women, and disadvantaged groups at risk. A study by UNESCO showed that possibly damaging stereotypes are the norm among AI chatbots. In 2019, Amazon scrapped its recruitment tool powered by artificial intelligence as the system was rigged against female applicants since previous hires were majorly male. Such bias leads to inequality and negative outcomes in recruitment, healthcare, education, and many more instances.
  2. Errors in facial recognition:
    It has been shown that even leading facial recognition technology can mismatch people. For example, in a study by the American Civil Liberties Union (ACLU) the software misidentified and mismatched 27 professional athletes to people in a database of criminals.
  3. Deepfakes:
    Deepfakes are artificial-intelligence-generated audio or video content with the intention to mislead. This technology can cause tremendous harm to society as it will contribute to the proliferation of cybercrime and misinformation. Deepfakes have already been used for selective social engineering attacks. In 2019, a journalist created a deepfake video of Mark Zuckerberg with a popular smartphone app.

Ethical AI benchmarks

Here are some moral concepts that should be common among the guidelines for ethical AI:

  1. Transparency:
    The decision-making system of the artificial intelligence machine must be transparent to users.
  2. Nonmaleficence:
    This is a term usually found in relation to the medical field it simply means “to do no harm”. The developers of AI-powered algorithms must ensure that the decisions taken by the system do not cause mental or physical harm to users.
  3. Justice:
    AI systems must be monitored regularly and closely to ensure that bias is not developed. The AI system should also have access to all genders and races to ensure equality.

How to build an ethical AI program

Here are some steps that can be used to create a customised, sustainable, and scalable ethical AI program:

  1. The first step is to identify any existing infrastructure that can be leveraged by the AI program. It is ideal to create an AI ethics program using the authority of existing infrastructure such as a governance board that meets to consider and review the privacy and data-related risks. If such an existing body does not already exist, companies should set up a dedicated committee for ethics-related issues.

  2. Companies should create a customised ethical framework that should include an elaboration of the company’s ethical standards. It should also establish relevant key performance indicators as well as a quality assurance program to appraise the effectiveness of the strategy. Moreover, the framework can show how ethical risk reduction is worked into business operations.

 

  1. Learn how industries such as healthcare approach AI ethically. Regulators, lawyers, medical professionals, and medical ethicists have explored the meaning of various topics such as informed consent, data privacy, and so on.

 

  1. Organisational awareness should be built and increased. There was a time when companies hardly paid attention to cybersecurity. Those times are over. Cybersecurity is important for all companies and all employees are expected to be aware of cybersecurity and the risks to cybersecurity. All departments and employees that come in contact with AI products should be made well aware of the company’s ethics framework and guidelines.

 

  1. Employees should be encouraged to identify AI ethical risks; this should be done both formally and informally. Encouraging employees can take the form of incentivising them via financial rewards.

  2. Companies need to keep a track of the impact of AI technology. Thorough research and review should be done to ensure that the product is used ethically. Stakeholders should be identified early on and they should be reviewed to find out how the product affects them.

The future of ethical AI

Google has called for governments around the world to increase the regulation of AI to prevent misuse and avoid mass surveillance programs and human rights violations. Today, there are several global initiatives committed to the cause of ethical AI. Regulation, framework, and guidelines can be used to ensure that AI is used for a safer world for everyone.

 Sources:

        https://www.itpro.co.uk/technology/30736/what-is-ethical-ai#:~:text=Roboethics%2C%20or%20robot%20ethics%2C%20is,the%20society%20it's%20operating%20in.

        https://hbr.org/2020/10/a-practical-guide-to-building-ethical-ai

        https://www.techslang.com/what-is-ethical-ai-and-why-we-need-to-talk-about-it/

        https://datafloq.com/read/we-need-ethical-ai-5-initiatives-ensure-ethical-ai/7571

Data points:

        https://en.unesco.org/Id-blush-if-I-could

        https://www.reuters.com/article/us-amazon-com-jobs-automation-insight/amazon-scraps-secret-ai-recruiting-tool-that-showed-bias-against-women-idUSKCN1MK08G

        https://www.aclum.org/en/news/facial-recognition-technology-falsely-identifies-famous-athletes/#athletes

        https://arstechnica.com/science/2019/12/how-i-created-a-deepfake-of-mark-zuckerberg-and-star-treks-data/

Wednesday 12 May 2021

Computer Vision: What is Computer Vision & AI Technologies

 Computer Vision

If you have watched an image trying to load on a social media app, you will usually see a description in its place. It will be a descriptor of things in the image such as a field of flowers or a child holding a balloon, and so on. This is computer vision. Facial recognition that is again the rage on social media apps and other software is another aspect of computer vision. Read on to find out more about this fascinating technology.

Computer vision

Computer vision or computer vision AI is a field of artificial intelligence where computers are trained to identify, interpret, and process images in videos and images in a similar way as humans do. Significant leaps in artificial intelligence, especially in the aspects of neural networks and deep learning have enabled computer vision AI to take great strides in recent years.

Computer vision uses

Computer vision AI is used by multiple industries ranging from retail to the military to healthcare and more. Here is a quick overview of how computer vision AI is used by different industries today:

  1. Retail:
    Retailers are using computer vision in a number of ways. They use it to increase loss prevention, uncover out-of-stock shelves, enhance shopping experiences, using self-checkout counters, and so on.

  2. Manufacturing:
    Here computer vision is being used to detect and identify manufacturing defects in real-time. During the assembly line process, the computer vision AI can detect even the smallest of defects on a product.

  3. Healthcare:
    Computer vision AI is being used to screen and thoroughly examine images from X-rays, CAT scans, and MRIs with a high rate of accuracy.

  4. Insurance:
    In the insurance industry, computer vision AI is being used to reduce fraud by enabling accurate and consistent assessments of vehicular damage.

  5. Defence and security:
    Computer vision is used extensively in the defence and security sectors. For example, casinos and banks use computer vision AI to analyse images from their security cameras. It is also used extensively to improve surveillance of power plants, embassies, hospitals, stadiums, railroads, and so on. It is also used to enhance the inspection of cargo at ports. Computer vision AI is also being used in reconnaissance missions by the military, as well as to quickly identify enemies, and to automate machine and vehicular movements.

  6. Agriculture:
    Computer vision has a tremendous impact on the agricultural industry. It allows farmers to detect diseases and pests quickly and on time. Early diagnosis prevents unnecessary loss and ensures product quality. Computer vision-enhanced robots monitor farms for weeds and spray herbicides to prevent or stop weeds from taking over. Furthermore, computer vision AI is used to help in the sorting of vegetables, fruits, and flowers by weight, size, quality, and other identifying markers.

  7. Autonomous vehicles:
    Computer vision AI
    plays an important role in self-driving vehicles. Here it plays a major role in perceiving and identifying the environment around the vehicle, thus allowing for navigation.

  8. Augmented reality:
    Here it detects objects in real-time and uses the information it has processed to place virtual objects within the same space.

  9. Facial recognition:
    This is where computer vision AI matches the images of people to their identities. This technology is not just restricted to security and defence sectors but is also used by social media apps and biometric authentication on smartphones, and so on.

Aspects of computer vision AI

There are four main aspects of computer vision AI also called the four eyes of computer vision. These are object recognition, video recognition, image recognition, and machine vision.

        Image recognition:
This is where computer vision is used to identify people, places, actions, and objects in an image.

        Object recognition:
This has a similar process to image recognition but it also allots a class label to which the image belongs. For example, in self-driving cars, computer vision recognises distinctions in traffic lights, and can also differentiate between a pedestrian and a lamppost.

        Video recognition:
This is where computer vision AI analyses video clips and compares those clips to a database of other content to ascertain if there is a match.

        Machine vision:
This is a hybrid of hardware and software. The technology here can conduct inspections and provide guidance to robots via visual feedback.

How does it work?

Here is a basic outline of how a computer vision model is built:

  1. A dataset is created incorporating annotated images. The annotation can be comprised of the image category, pixel-wise segmentation of the objects present, or pairs of classes and bounding boxes.
  2. From each image, features that are relevant to the task are extracted. For example, in images, for a task to recognise people, the computer vision AI will recognise features based on facial features.
  3. Then a deep learning model is trained based on the features that are isolated in the dataset. This means that the AI is fed many more images to learn how to solve the required task.
  4. Thereafter, the computer vision model is evaluated by using images that were not used during the training phase.

Limitations of computer vision

        The need for specialists: Companies need a team of highly skilled specialists that can build and use computer vision AI.

        The need for regular monitoring: To prevent technical glitches, companies need a dedicated team to constantly monitor and evaluate the system

        No replacement for human intelligence and vision: Computer vision AI has come a long way but it is not a replacement for human intelligence and human vision. Machines do not understand the complexities and intricacies of the world around us the way humans can.

The bottom line

Computer vision AI is an exciting technology that has taken the field of artificial intelligence to new heights. The world will surely see much more advancements and innovations in this field in the future and it will continue to revolutionise our world as we know it.

Sources:

  1. https://towardsdatascience.com/an-overview-of-computer-vision-1f75c2ab1b66
  2. https://www.sas.com/en_in/insights/analytics/computer-vision.html#defense-security
  3. https://tryolabs.com/resources/introductory-guide-computer-vision/
  4. https://xd.adobe.com/ideas/principles/emerging-technology/what-is-computer-vision-how-does-it-work/
  5. https://www.bbntimes.com/technology/benefits-and-limitations-of-computer-vision
Related Posts Plugin for WordPress, Blogger...

Share Buttons