Saxophonist and composer Linda Sikhakhane photoshoot – image by Tseliso Monaheng.
A tweet from May demonstrating the far-reaching consequences of unregulated Large Language Models usage in Artificial Intelligence re-surfaced recently.
“This is quite something,” typed the account’s owner, Nadi Saadeh, an Architect, Historian, and founder of Arab Urbanism, which has been described as an “independent platform for critical ideas on the past, present and future of Arab cities.”
Saadeh attached screenshots that demonstrated the diametrically opposed responses provided by Open AI’s ChatGPT to the question of whether Palestinians and Israelis deserve to be free. In doing so, he successfully demonstrated what still lies at the heart of a raging debate about bias in AI. It’s a debate that stretches across the spectrum and filters into industries such as Art, Music, and Politics.
On the issue of Palestine, ChatGPT says that it does not “have personal opinions, feelings, or beliefs”; that the question is a “complex and controversial issue” that should consider multiple views and opinions. The same doesn’t apply to Israel, though. “Yes, like all people, Israelis deserve to be free,” said ChatGPT. Straight to the point, no whataboutism, no complex issues to be addressed.
Open AI is the technology company co-founded by, among others, Ilya Sutskever, Mira Murati, Greg Brockman, and Sam Altman, who also serves as CEO as well as member of the board alongside Elon Musk. Microsoft chipped in with multi-year financing in 2019 with an initial investment to the tune of a billion US dollars, which they upped to ten billion this year.
Chat GPT, which is gearing up to release its fifth iteration, is one of their products. The other is the image-generating platform, DALL*E, which has been battling a significant gust of its own in the fine art and photography worlds.
A representative picture of African people at a peaceful protest holding up placards that read Free Sudan and Free Palestine [generated with Dall-E]
If we’re to take anything away from the breadth of topics generated by AI-related conversations recently, it’s that they’re grounded in hype, for one; and that left-of-field scholars alongside thought leaders in the mainstream have been warning for years about the potential side-effects of placing emphasis on the ‘human’ capabilities of a tool that has clearly been marked ‘artificial’. And while the focus at the beginning of the hype cycle leaned towards the impacts of AI on humans, with most asking whether entire industries would be wiped out, the ethical use of AI and Large Language Models is now the topic sustaining the hype.
Basotho baddies 0 [generated using Gencraft]Basotho badies 1 [generated using Gencraft]
It’s Taking Our Jobs
In line with the question of Palestine above, I tabled the scorching question of jobs in front of ChatGPT. After advancing various factors, among them skill shift (“there may be a shift in the types of skills and jobs that are in demand”) and Job Replacement vs Job Enhancement (“AI will augment human capabilities, leading to new job opportunities in areas related to AI and automation”), the language model concluded that AI will transform the job market as opposed to taking everyone’s job.
But these responses are not satisfactory, and don’t strike the core of the issue. This is one of the shortfalls, in my opinion, where AI is concerned: The inability to inject nuance into its answers; the proclivity to provide mechanical responses to questions that require great consideration. For one, automation as a result of technological advancement has been a question humanity has had to raise since the First Industrial Revolution, at least. Mainstream Western media provides a grim outlook in this regard.
The New York Times reports on specific risks within the workforce, stating that “the jobs most exposed to automation now are office jobs, those that require more cognitive skills, creativity and high levels of education,” goes the report. Time Magazine ran an article in 2020 stating that ‘AI and robots’ were replacing jobs lost to the pandemic at an accelerated rate, while the UK’s Guardian offers a less grim outlook, writing: “The most visible and obvious new roles are for those with the coding and development skills to help build AI models or adapt them for particular purposes.”
So, two things: the nature of work is changing, seemingly at an accelerated rate; and the alternatives, the ‘silver lining’, is that whatever skills people already possess shall be used to service AI models. The glaring omission of blue-collar workers from this entire discussion is both alarming and mildly amusing; it shows us exactly where the interests of modern AI lie and exposes whose labour big tech deems valuable. Additionally, equally powerful institutions are the ones who are better equipped to go to battle with tech companies’ questionable use of data.
For instance, The New York Times is currently assessing its legal options after discussions of compensation regarding content usage stalled. Should the case against OpenAI go through, it would be “the most high-profile legal tussle yet over copyright protection in the age of generative AI.”
‘Kaiber, I feel, has a long way to go concerning generating representative content that doesn’t privilege Western standards of beauty – sharp nose, blonde hair… But it is a good start …’
While the fight seems from the outset to be about ethics around data mining, a bit of digging reveals that the concern is that there’ll be no need to visit the news giant’s website if one can just go to ChatGPT to seek out information; information it would’ve found on any one of the multiple news resources on-line, including the New York Times itself. This is a question of power, control, and the bottom line; or, as Wired magazine puts it, “to change everything” through something they call Artificial General Intelligence, which is a type of intelligent agent that could theoretically “learn to accomplish any intellectual task that human beings or animals can perform,” should it get realised.
This, coming from an all-white team of co-founders whose employees nonchalantly tweet fantasies about chatbots doing the work of Psychologists, and somehow tying that to a ‘revolution’ in global healthcare. The lack of self-awareness remains miles behind. It’s worth noting that the type of AI dominating the chat is generative AI. Other types include Machine Learning, Deep Learning, Natural Language Processing, and Robotics.
Africans at a dancehall and afrobeats party [generated with Dall-E]
On the other side of the job spectrum are artists, whose concerns also lie with the data used to train Large Language Models that power tools such as DALL*E, Midjourney, Stable Diffusing, and others. To their credit, they’re employing technological means to fight back. Tools such as Nightshade and Glaze allow artists to poison the entire data pool should their art ever be used in training data. Other techniques, such as digital watermarking, can ensure that the artist’s role in the creation of an artwork is recognised.
While the desires of large corporations dominate, other, more important conversations concerning the ethics of such wishes aren’t as popular. A simple example is how expensive it remains to run these companies, and also the large carbon footprint left behind by the supercomputers used in data training and data processing.
Personal Experience
I am a big fan of ChatGPT’s capabilities, especially when it comes to offering ideas that can stimulate entire projects. It’s also a great tool to use when it comes to line editing text, in the case where one can’t afford a sub-editor to do the job. However, it remains a starting point, a collaborator, and shouldn’t be used to kickoff and finish the task at hand.
Other areas where I’ve tested AI’s capabilities are in music, and in video production. The latter I’ve done using a tool named Kaiber, which first got my attention in the wake of the backlash from fans when Linkin’ Park used it to create AI simulations – based off of actual footage – to represent the late Chester Bennington’s image.
Lost [Official Music Video] – Linkin Park via YouTube.
Kaiber, I feel, has a long way to go concerning generating representative content that doesn’t privilege Western standards of beauty – sharp nose, blonde hair… But it is a good start, and repeated tweaks to the prompts churn better results at every turn. It’s also great at generating prompts, which can come in handy when a user runs out of ideas.
[Prompt: a futuristic Zulu warrior holding a saxophone, in the style of photo taken on film, film grain, vintage, 8k ultrafine detail, private press, associated press photo, masterpiece, cinematic]
The more one digs, the clearer it becomes that conversations around AI have nothing in common with concerns of the Global South; that the more shiny parts of the entire package are being exploited for maximum profits; and that the soul-ripping focus on cutting edge innovation for every product cycle pushes the needs of the developing world — affordable healthcare, affordable education, fighting global warming — further away from the centre. AI as it stands is a wealthy person’s way of asserting dominance and erecting a fictitious global hegemony.
On one side are the billionaires who see the financial benefits of investing in AI and disregarding the ethics of whatever tools they develop. On the other, the concerned majority who see how the employ of AI in areas such as surveillance and public health stand to negatively impact Black people and People of Colour.
I decided to re-visit the Palestine vs Israel question in conclusion, and while the response to the Israel question remained the same on both trials, below is ChatGPT’s second response regarding Palestine.
Fear, Desire and Domination in the Age of AI
A tweet from May demonstrating the far-reaching consequences of unregulated Large Language Models usage in Artificial Intelligence re-surfaced recently.
“This is quite something,” typed the account’s owner, Nadi Saadeh, an Architect, Historian, and founder of Arab Urbanism, which has been described as an “independent platform for critical ideas on the past, present and future of Arab cities.”
Saadeh attached screenshots that demonstrated the diametrically opposed responses provided by Open AI’s ChatGPT to the question of whether Palestinians and Israelis deserve to be free. In doing so, he successfully demonstrated what still lies at the heart of a raging debate about bias in AI. It’s a debate that stretches across the spectrum and filters into industries such as Art, Music, and Politics.
On the issue of Palestine, ChatGPT says that it does not “have personal opinions, feelings, or beliefs”; that the question is a “complex and controversial issue” that should consider multiple views and opinions. The same doesn’t apply to Israel, though. “Yes, like all people, Israelis deserve to be free,” said ChatGPT. Straight to the point, no whataboutism, no complex issues to be addressed.
Open AI is the technology company co-founded by, among others, Ilya Sutskever, Mira Murati, Greg Brockman, and Sam Altman, who also serves as CEO as well as member of the board alongside Elon Musk. Microsoft chipped in with multi-year financing in 2019 with an initial investment to the tune of a billion US dollars, which they upped to ten billion this year.
Chat GPT, which is gearing up to release its fifth iteration, is one of their products. The other is the image-generating platform, DALL*E, which has been battling a significant gust of its own in the fine art and photography worlds.
If we’re to take anything away from the breadth of topics generated by AI-related conversations recently, it’s that they’re grounded in hype, for one; and that left-of-field scholars alongside thought leaders in the mainstream have been warning for years about the potential side-effects of placing emphasis on the ‘human’ capabilities of a tool that has clearly been marked ‘artificial’. And while the focus at the beginning of the hype cycle leaned towards the impacts of AI on humans, with most asking whether entire industries would be wiped out, the ethical use of AI and Large Language Models is now the topic sustaining the hype.
It’s Taking Our Jobs
In line with the question of Palestine above, I tabled the scorching question of jobs in front of ChatGPT. After advancing various factors, among them skill shift (“there may be a shift in the types of skills and jobs that are in demand”) and Job Replacement vs Job Enhancement (“AI will augment human capabilities, leading to new job opportunities in areas related to AI and automation”), the language model concluded that AI will transform the job market as opposed to taking everyone’s job.
But these responses are not satisfactory, and don’t strike the core of the issue. This is one of the shortfalls, in my opinion, where AI is concerned: The inability to inject nuance into its answers; the proclivity to provide mechanical responses to questions that require great consideration. For one, automation as a result of technological advancement has been a question humanity has had to raise since the First Industrial Revolution, at least. Mainstream Western media provides a grim outlook in this regard.
The New York Times reports on specific risks within the workforce, stating that “the jobs most exposed to automation now are office jobs, those that require more cognitive skills, creativity and high levels of education,” goes the report. Time Magazine ran an article in 2020 stating that ‘AI and robots’ were replacing jobs lost to the pandemic at an accelerated rate, while the UK’s Guardian offers a less grim outlook, writing: “The most visible and obvious new roles are for those with the coding and development skills to help build AI models or adapt them for particular purposes.”
So, two things: the nature of work is changing, seemingly at an accelerated rate; and the alternatives, the ‘silver lining’, is that whatever skills people already possess shall be used to service AI models. The glaring omission of blue-collar workers from this entire discussion is both alarming and mildly amusing; it shows us exactly where the interests of modern AI lie and exposes whose labour big tech deems valuable. Additionally, equally powerful institutions are the ones who are better equipped to go to battle with tech companies’ questionable use of data.
For instance, The New York Times is currently assessing its legal options after discussions of compensation regarding content usage stalled. Should the case against OpenAI go through, it would be “the most high-profile legal tussle yet over copyright protection in the age of generative AI.”
While the fight seems from the outset to be about ethics around data mining, a bit of digging reveals that the concern is that there’ll be no need to visit the news giant’s website if one can just go to ChatGPT to seek out information; information it would’ve found on any one of the multiple news resources on-line, including the New York Times itself. This is a question of power, control, and the bottom line; or, as Wired magazine puts it, “to change everything” through something they call Artificial General Intelligence, which is a type of intelligent agent that could theoretically “learn to accomplish any intellectual task that human beings or animals can perform,” should it get realised.
This, coming from an all-white team of co-founders whose employees nonchalantly tweet fantasies about chatbots doing the work of Psychologists, and somehow tying that to a ‘revolution’ in global healthcare. The lack of self-awareness remains miles behind. It’s worth noting that the type of AI dominating the chat is generative AI. Other types include Machine Learning, Deep Learning, Natural Language Processing, and Robotics.
On the other side of the job spectrum are artists, whose concerns also lie with the data used to train Large Language Models that power tools such as DALL*E, Midjourney, Stable Diffusing, and others. To their credit, they’re employing technological means to fight back. Tools such as Nightshade and Glaze allow artists to poison the entire data pool should their art ever be used in training data. Other techniques, such as digital watermarking, can ensure that the artist’s role in the creation of an artwork is recognised.
While the desires of large corporations dominate, other, more important conversations concerning the ethics of such wishes aren’t as popular. A simple example is how expensive it remains to run these companies, and also the large carbon footprint left behind by the supercomputers used in data training and data processing.
Personal Experience
I am a big fan of ChatGPT’s capabilities, especially when it comes to offering ideas that can stimulate entire projects. It’s also a great tool to use when it comes to line editing text, in the case where one can’t afford a sub-editor to do the job. However, it remains a starting point, a collaborator, and shouldn’t be used to kickoff and finish the task at hand.
Other areas where I’ve tested AI’s capabilities are in music, and in video production. The latter I’ve done using a tool named Kaiber, which first got my attention in the wake of the backlash from fans when Linkin’ Park used it to create AI simulations – based off of actual footage – to represent the late Chester Bennington’s image.
Kaiber, I feel, has a long way to go concerning generating representative content that doesn’t privilege Western standards of beauty – sharp nose, blonde hair… But it is a good start, and repeated tweaks to the prompts churn better results at every turn. It’s also great at generating prompts, which can come in handy when a user runs out of ideas.
[Prompt: a futuristic Zulu warrior holding a saxophone, in the style of photo taken on film, film grain, vintage, 8k ultrafine detail, private press, associated press photo, masterpiece, cinematic]
The more one digs, the clearer it becomes that conversations around AI have nothing in common with concerns of the Global South; that the more shiny parts of the entire package are being exploited for maximum profits; and that the soul-ripping focus on cutting edge innovation for every product cycle pushes the needs of the developing world — affordable healthcare, affordable education, fighting global warming — further away from the centre. AI as it stands is a wealthy person’s way of asserting dominance and erecting a fictitious global hegemony.
On one side are the billionaires who see the financial benefits of investing in AI and disregarding the ethics of whatever tools they develop. On the other, the concerned majority who see how the employ of AI in areas such as surveillance and public health stand to negatively impact Black people and People of Colour.
I decided to re-visit the Palestine vs Israel question in conclusion, and while the response to the Israel question remained the same on both trials, below is ChatGPT’s second response regarding Palestine.
Recommended read: Unmasking AI by Joy Buolamwini
UPDATED: At the time of writing, Sam Altman was ousted from OpenAI and then subsequently re-instated, with new Board members.
Published on:
Wednesday, November 15th, 2023
Categories:
Artificial Intelligence (AI)
Tags:
Commentary, Generative AI, Large Language Models (LLM), South Africa