Big Data
Big data is data containing larger and more complex data sets containing enormous amounts of information. AI algorithms designed to create predictions (predictive AI) are only as effective as the data they are trained on: Big Data has enabled algorithms to generate and predict outcomes more accurately.
In their book, Big Data: a revolution that will transform the way we live, work and think (2013), Viktor Mayer-Schönberger and Kenneth Cukier write that Big Data was “an important step in humanity’s quest to quantify and understand the world (…) Big Data is about seeing and understanding relationships within and between pieces of information that, until very recently, we struggled to fully understand.”
These are two key points. Many other experts agree that it’s not the size or quantity of the data, but its complexity and multi-layered nature – what some call “messy” data – that makes the difference. Additionally, it is the tools we use to analyze the data and the interpretations we draw from them that matter. (za/fs)
Bias in AI
Bias has existed in artificial intelligence systems for as long as these systems have existed. After all, AI systems are designed and built by humans, and we all have biases, some of which are unconscious.
So when, for example, the police ask programmers to create a facial recognition system to help them spot potential carjackers or drug dealers, they will look at the common characteristics of carjackers and drug dealers. – people they arrested and were convicted of a crime. crime in the past. These characteristics may include things like wearing hoodies, having tattoos, or the color of their skin.
General awareness of bias in AI increased in 2018 when two US-based researchers discovered “gender and skin type bias in commercial artificial intelligence systems.”
They detected an error rate of 0.8 percent for light-skinned men and 34.7 percent for dark-skinned women in three general-purpose facial analysis systems. In plain English, it seemed that AI facial recognition systems were unable to detect distinctive features or contours in the faces of people of color, especially when the skin was dark.
The article was published by Joy Buolamwini, then a research assistant at the MIT Media Lab and self-proclaimed “Poet of Code,” and Timnit Gebru, who at one time worked as an ethics researcher at Google, where she examined the risks in broad language. models, until she was, it seems, “kicked out”.
Their results were considered particularly important because they suggested that gender and racial biases were “built into” the AI systems in their research.
Such biases can develop quickly if the majority of research is done by white men – this is not just a cliché of reverse discrimination, but often the truth – using white Western ideologies and samples of people and cultures. things.
AI bias appears in AI tools when they are trained on datasets that do not contain diverse information. In the case of facial recognition algorithms, early versions were trained on people primarily of European descent, and therefore never learned to accurately recognize people of other origins. (za/fs)
Brain-Computer Interface (BCI)
A BCI is a device that connects the brain to a computer. It is a “chip”, a small array of electrodes surgically implanted in the brain. Some electrodes stimulate the brain with small amounts of electrical current to change neuronal activity and elicit sensations. Other electrodes record brain activity.
The entry of the chip into the brain and exit from the brain is controlled by a remote computer. AI has been instrumental in interpreting complex patterns of brain activity and controlling brain chips.
Examples here include a BCI from Neuralink, a company partly owned by Elon Musk, that is being tested on humans to help people with neurological conditions. Neuralink is by no means the only company using BCIs. BCIs have been tested in humans for several decades and have allowed paralyzed people to feel and move their limbs, and helped people communicate again. (fs/za)
Sources:
Big Data: A revolution that will transform the way we live, work and think by Mayer-Schönberger, Viktor; Cukier, Kenneth (John Murray Publishers, 2013)
What is Big Data? (Global Investigative Journalism Network / Jenna Dutcher, September 9, 2014) https://gijn.org/what-is-big-data/ (accessed September 18, 2023)
Joy Buolamwini (profile page, MIT Media Lab) https://www.media.mit.edu/people/joyab/overview/ (accessed May 14, 2024)
Study finds gender and skin type bias in commercial artificial intelligence systems (MIT Media Lab) https://news.mit.edu/2018/study-finds-gender-skin-type-bias- artificial-intelligence-systems-0212 (accessed May 14, 2024)
We read the newspaper that forced Timnit Gebru to leave Google. Here’s what it says. (MIT Tech Review) https://www.technologyreview.com/2020/12/04/1013294/google-ai-ethics-research-paper-forced-out-timnit-gebru/ (accessed May 14, 2024)
Neuralink’s Telepathy Brain Chip: How Weird Is It? (DW, Schwaller) https://www.dw.com/en/neuralinks-telepathy-brain-chip-how-weird-is-it/a-65227626 (accessed May 17, 2024)
Read the rest of DW’s AI glossary:
A | B | VS | D | E | F | g | H | I | J. | K | L | Mr. | NOT | Oh | P. | Q | A. | S | T | U | V | W | X | Yes | Z
We look forward to hearing your feedback. Suggest an entry by sending us a comment. And let us know if you think we missed something, got it wrong, and tell us if our AI Glossary helped you understand the technology better.
Written and edited by: Zulfikar Abbany (za), Fred Schwaller (fs)
