AI (124)
Find narratives by ethical themes or by technologies.
FILTERreset filters-
- 10 min
- MIT Technology Review
- 2020
This article explains the ethical warnings of Timnit Gebru against training Natural Language Processing algorithms on large language models developed on sets of textual data from the internet. Not only does this process have a negative environmental impact, it also still does not allow these machine learning tools to process semantic nuance, especially as it relates to burgeoning social movements or countries with lower internet access. Dr. Gebru’s refusal to retract this paper ultimately lead to her dismissal from Google.
- MIT Technology Review
- 2020
-
- 10 min
- MIT Technology Review
- 2020
We read the paper that forced Timnit Gebru out of Google. Here’s what it says.
This article explains the ethical warnings of Timnit Gebru against training Natural Language Processing algorithms on large language models developed on sets of textual data from the internet. Not only does this process have a negative environmental impact, it also still does not allow these machine learning tools to process semantic nuance, especially as it relates to burgeoning social movements or countries with lower internet access. Dr. Gebru’s refusal to retract this paper ultimately lead to her dismissal from Google.
How should models for training NLP algorithms be more closely scrutinized? What sorts of voices are needed at the design table to ensure that the impact of such algorithms are consistent across all populations? Can this ever be achieved?
-
- 7 min
- VentureBeat
- 2021
The GPT-3 Natural Language Processing model, created by the company open AI and released in 2020, is the most powerful of its kind, using a generalized approach to feed its machine learning algorithm in order to mirror human speech. The potential applications of such a powerful program are manifold, but this potential means that many tech monopolies may want to enter an “arms race” to get the most powerful model possible.
- VentureBeat
- 2021
-
- 7 min
- VentureBeat
- 2021
GPT-3: We’re at the very beginning of a new app ecosystem
The GPT-3 Natural Language Processing model, created by the company open AI and released in 2020, is the most powerful of its kind, using a generalized approach to feed its machine learning algorithm in order to mirror human speech. The potential applications of such a powerful program are manifold, but this potential means that many tech monopolies may want to enter an “arms race” to get the most powerful model possible.
Should AI be able to imitate human speech unchecked? Should humans be trained to be able to tell when speech or text might be produced by a machine? How might Natural Language Processing cheapen human writing and writing jobs?
-
- 5 min
- Gizmodo
- 2021
Thorough investigation led to the conclusion that bots played a role in the economic disruption of GameStop stocks in early 2021. Essentially, the automated accounts aided in the diffusion of materials promoting the purchase and maintenance of GameStop stocks as a ploy to act as a check on wealthy hedge fund managers who bet that the stock would crash. The wholistic effect of these bots in this specific campaign, and thus a measure of how bots may generally be used to cause economic disruption in online markets through interaction with humans, remains hard to read.
- Gizmodo
- 2021
-
- 5 min
- Gizmodo
- 2021
Bots Reportedly Helped Fuel GameStonks Hype on Facebook, Twitter, and Other Platforms
Thorough investigation led to the conclusion that bots played a role in the economic disruption of GameStop stocks in early 2021. Essentially, the automated accounts aided in the diffusion of materials promoting the purchase and maintenance of GameStop stocks as a ploy to act as a check on wealthy hedge fund managers who bet that the stock would crash. The wholistic effect of these bots in this specific campaign, and thus a measure of how bots may generally be used to cause economic disruption in online markets through interaction with humans, remains hard to read.
Do you consider this case study, and the use of the bots, to be “activism”? How can this case study be summarized into a general principle for how bots may manipulate the economy? How do digital technologies help both wealth and non-wealthy people serve their own interests?
-
- 7 min
- New York Times
- 2018
This article details the research of Joy Buolamwini on racial bias coded into algorithms, specifically facial recognition programs. When auditing facial recognition software from several large companies such as IBM and Face++, she found that they are far worse at properly identifying darker skinned faces. Overall, this reveals that facial analysis and recognition programs are in need of exterior systems of accountability.
- New York Times
- 2018
-
- 7 min
- New York Times
- 2018
Facial Recognition Is Accurate, if You’re a White Guy
This article details the research of Joy Buolamwini on racial bias coded into algorithms, specifically facial recognition programs. When auditing facial recognition software from several large companies such as IBM and Face++, she found that they are far worse at properly identifying darker skinned faces. Overall, this reveals that facial analysis and recognition programs are in need of exterior systems of accountability.
What does exterior accountability for facial recognition software look like, and what should it look like? How and why does racial bias get coded into technology, whether explicitly or implicitly?
-
- 7 min
- The Verge
- 2020
PULSE is an algorithm which can supposedly determine what a face looks like from a pixelated image. The problem: more often than not, the algorithm will return a white face, even when the person from the pixelated photograph is a person of color. The algorithm works through creating a synthetic face which matches with the pixel pattern, rather than actually clearing up the image. It is these synthetic faces that demonstrate a clear bias toward white people, demonstrating how institutional racism makes its way thoroughly into technological design. Thus, diversity in data sets will not full help until broader solutions combatting bias are enacted.
- The Verge
- 2020
-
- 7 min
- The Verge
- 2020
What a machine learning tool that turns Obama white can (and can’t) tell us about AI bias
PULSE is an algorithm which can supposedly determine what a face looks like from a pixelated image. The problem: more often than not, the algorithm will return a white face, even when the person from the pixelated photograph is a person of color. The algorithm works through creating a synthetic face which matches with the pixel pattern, rather than actually clearing up the image. It is these synthetic faces that demonstrate a clear bias toward white people, demonstrating how institutional racism makes its way thoroughly into technological design. Thus, diversity in data sets will not full help until broader solutions combatting bias are enacted.
What potential harms could you see from the misapplication of the PULSE algorithm? What sorts of bias-mitigating solutions besides more diverse data sets could you envision? Based on this case study, what sorts of real-world applications should facial recognition technology be trusted with?
-
- 10 min
- Gizmodo
- 2021
Physicist Brian Nord, who learned about deep learning algorithms through his research on the cosmos, warns against how developing algorithms without proper ethical sensibility can lead to these algorithms having more negative impacts than positive ones. Essentially, an “a priori” or proactive approach to instilling AI ethical sensibility, whether through review institutions or ethical education of developers, is needed to guard against privileged populations using algorithms to maintain hegemony.
- Gizmodo
- 2021
-
- 10 min
- Gizmodo
- 2021
Developing Algorithms That Might One Day Be Used Against You
Physicist Brian Nord, who learned about deep learning algorithms through his research on the cosmos, warns against how developing algorithms without proper ethical sensibility can lead to these algorithms having more negative impacts than positive ones. Essentially, an “a priori” or proactive approach to instilling AI ethical sensibility, whether through review institutions or ethical education of developers, is needed to guard against privileged populations using algorithms to maintain hegemony.
What would an ideal algorithmic accountability organization or process look like? What specific ethical regions should AI developers study before creating their algorithms? How can algorithms or other programs created for one context, such as scientific research or learning, be misused in other contexts?