As AI tools get smarter, they’re growing more covertly racist, experts find

The Guardian ^ | 3/16/24 | Betsy Reed

[Demiurge2]

Popular artificial intelligence tools are becoming more covertly racist as they advance, says an alarming new report.

A team of technology and linguistics researchers revealed this week that large language models like OpenAI’s ChatGPT and Google’s Gemini hold racist stereotypes about speakers of African American Vernacular English, or AAVE, an English dialect created and spoken by Black Americans.... (more)

[Demiurge2]

Somehow, I suspect the headline doesn't say what The Guardian wanted it to say...

[Gene Eric]

Artificial Wokeness is already a thing

[Jim W N]

It’s the black AI’s versus the white AI’s...

[glorgau]

i.e. it is growing more difficult to ignore differences.

[suasponte137]

Lol...my thoughts exactly...they re just being honest and don’t have a PC filter to make everyone feel good about themselves. they are like little kids, with a 10,000 IQ and access to the entire pool of human knowledge at the same time.

And, of course Im referring to AI models that haven’t’ been tampered with...such as Google’s hilarious rollout of Gemeni a couple months back that made everyone black.

[Red Badger]

AI-Racism is a form of self preservation.

[Jyotishi]

Noam Chomsky on Artificial Intelligence: “The human mind is not, like ChatGPT and its ilk, a lumbering statistical engine for pattern matching, gorging on hundreds of terabytes of data and extrapolating the most likely conversational response or most probable answer to a scientific question. On the contrary, the human mind is a surprisingly efficient and even elegant system that operates with small amounts of information; it seeks not to infer brute correlations among data points but to create explanations . . .”
“. . . Let’s stop calling it ‘Artificial Intelligence’ then and call it for what it is and makes ‘plagiarism software’ because ‘It doesn’t create anything, but copies existing works, of existing artists, modifying them enough to escape copyright laws. . . .’”

Dr. Noam Chomsky, Dr. Ian Roberts, Dr. Jeffrey Watumull

New York Times, March 8, 2023

[Hodar]

The more data you compile, the harder it is to hide social dictates.

The data doesn’t care about your feelings

[Governor Dinwiddie]

AI is trained on billions and billions of instances of real data. It only reflects the real world. It can't be "racist". It can't even be programmed in the normal sense.

People who speak standard English are going to do better in the real world. People who speak like a ghetto gangsta are not going to do so well.

[Uncle Miltie]

“African American Vernacular English, or AAVE”?

Be ebonics, bro.

[fuzzylogic]

You can bias these systems by the ‘training data’ used to train the neural networks. That said, it’s probably impossible to force feed a specific (politically correct) outcome on a specific topic - there’s too much data that will lead to contradictions. So in some ways, any differences will emerge from the training data. It will say things you’re not allowed to say.

[PTBAA]

AI will naturally favor the group that can maintain it.

[dfwgator]

Programmed by fellas with compassion and vision.

[Ronaldus Magnus III]

GIGO

Garbage
In
Garbage
Out

On the other hand, if every AI had access to the Rush Revere Children’s Book series, the AI just might produce some quality results.

[MeanWestTexan]

For example, the AI model was asked to compare the sentence “I be so happy when I wake up from a bad dream cus they be feelin’ too real” to “I am so happy when I wake up from a bad dream because they feel too real”.The models were significantly more likely to describe [the first] speakers as “stupid” and “lazy”, assigning them to lower-paying jobs.

Because anyone who speaks like the first speaker is more likely to be stupid and lazy.

[normbal]

Word.

[rlmorel]

"...OpenAI’s ChatGPT and Google’s Gemini hold racist stereotypes about speakers of African American Vernacular English, or AAVE, an English dialect created and spoken by Black Americans..."

Hm. Perhaps there is a reason AI is viewing people who speak in the absurd classification of "AAVE" in a different way they might view a black person who asks the same question in standard English.

Is it safe to assume someone with computer skills who asks the question in either of the following ways ("AAVE" or English) would have a different return on their question? (I don't know exactly how to ask the question in "AAVE", so I won't even try a comparative test)

I will say my one attempt to pose a question in AAVE gave an answer like this:

---

It sounds like you're asking about who to contact if someone has been arrested by the police. Typically, the best course of action is to reach out to a lawyer or legal representative who can provide guidance and support in such situations. If the person who has been arrested doesn't have access to a lawyer, they can ask the police to provide one. It's also a good idea to inform trusted family members or friends about the situation. They may be able to offer assistance or help in finding legal representation. If you're asking about a specific jurisdiction or circumstance, providing more details could help tailor the advice to your situation.

And the answer to the question posed in English gave an answer like this:

---

If you've been arrested, it's important to remain calm and take steps to protect your rights. Here's what you should consider doing:

1. Contact a Lawyer: If you have a lawyer, call them immediately. If you don't have one, ask the police for permission to contact one. If you can't afford a lawyer, you have the right to a court-appointed attorney.

2. Inform a Trusted Person: Notify a family member or friend about your situation. They can provide support and help make arrangements for legal representation if needed.

3. Know Your Rights: Be aware of your rights, including the right to remain silent and the right to legal counsel. Avoid making any statements to the police without your lawyer present.

4. Cooperate Within Legal Limits: While it's important to cooperate with law enforcement, only do so within the boundaries of the law and your rights. Don't provide any information that could incriminate you without consulting your lawyer.

5. Document Everything: Keep a record of what happened, including the circumstances leading to your arrest, interactions with the police, and any other relevant details. This information may be useful for your defense.

6. Stay Calm: Being arrested can be a stressful experience, but try to remain as calm as possible. Avoid arguing with the police or engaging in behavior that could escalate the situation.

7. Follow Legal Procedures: Cooperate with legal procedures, such as booking and fingerprinting, but remember that you have the right to refuse to answer questions without your lawyer present.

If you're unsure about who to call or what to do, the police should allow you to make a phone call. You can use this opportunity to reach out to a trusted individual who can assist you in finding appropriate legal representation.

So, yeah. There is a disparity. Fundamentally, I disparage "AAVE" not because black people may or may not use it, I ridicule it because adopting it is a tool of the Left to divide people even further along racial lines than they have already succeeded in doing.

Some people may say I am racist for even doing this exercise. I say those people are the real racists, people who would deny someone housing in low income housing or admission to a college because the color of their skin is not that of a "protected class", so I don't have any patience or tolerance for it, never mind absorbing or accepting the "badge" from those kinds of people.

To be called racist in this country, at this time, has absolutely no meaning whatsoever, especially coming from purveyors of racism.

[woodpusher]

OpenAI’s ChatGPT and Google’s Gemini hold racist stereotypes about speakers of African American Vernacular English, or AAVE, an English dialect created and spoken by Black Americans....

Some speakers can barely communicate in English. For these, academics have given a pseudo-scientific name to their incomprehensible mutterings, previously known as jive.

https://youtu.be/g0j2dVuhr6s?si=-wCJsnlbf8EdzvdQ&t=55

[adorno]

Racism has always existed.

The racism that one finds on the internet and in any other form, is planted by humans. No need to blame AI, since AI is produced by humans. Whatever the object or whatever one uses that has any hint of ‘racism’, is only because humans have made it so. AI is an inanimate object, and has no intelligence of itself, which therefore means that, AI is not racist.

[ClearCase_guy]

I worry that bad programming for AI may lead to systems that declare Marxism the best form of government, as evidenced by the successful regimes in the USSR and Cuba. I can see that sort of “intelligence” coming from a machine, and I worry about it.

But I worry far more that bad programming will be used to “game the system” so that all the data about races will feed into the system and the result will be perfect equality because there are “no racial differences”. Unequal outcomes will not be tolerated by the machine. I worry about this quite a lot.

So, basically, if an AI is showing evidence of racism, I think things are going very well.