openai – Techdirt (original) (raw)

How Refugee Applications Are Being Lost In (Machine) Translation

from the AI-not-I dept

As you may have noticed, headlines are full of the wonders of chatbots and generative AI these days. Although often presented as huge breakthroughs, in many ways they build on machine learning techniques that have been around for years. These older systems have been deployed in real-life situations for some time, which means they provide valuable information about the possible pitfalls of using AI for serious tasks. Here is a typical example of what has been happening in the world of machine translation when applied to refugee applications for asylum, as reported on the Rest of the World site:

A crisis translator specializing in Afghan languages, Mirkhail was working with a Pashto-speaking refugee who had fled Afghanistan. A U.S. court had denied the refugee’s asylum bid because her written application didn’t match the story told in the initial interviews.

In the interviews, the refugee had first maintained that she’d made it through one particular event alone, but the written statement seemed to reference other people with her at the time — a discrepancy large enough for a judge to reject her asylum claim.

After Mirkhail went over the documents, she saw what had gone wrong: An automated translation tool had swapped the “I” pronouns in the woman’s statement to “we.”

That’s a tiny difference, and one that today’s machine translation programs can easily miss, especially for languages where training materials are still scarce. And yet the consequences of the shift from singular “I” to plural “we” can have life-changing consequences – in the case above, whether asylum was granted to a refugee fleeing Afghanistan. There are other problems too:

Based in New York, the Refugee Translation Project works extensively with Afghan refugees, translating police reports, news clippings, and personal testimonies to bolster claims that asylum seekers have a credible fear of persecution. When machine translation is used to draft these documents, cultural blind spots and failures to understand regional colloquialisms can introduce inaccuracies. These errors can compromise claims in the rigorous review so many Afghan refugees experience.

In the future it is likely that the number of people seeking asylum will increase, not least because of environmental refugees who are fleeing lands made uninhabitable by climate change. Their applications for asylum elsewhere are likely to involve a wider range of lesser-known languages. Turning to machine translation will be a natural move by the authorities, since it takes time and resources to recruit specialist human translators.

The new generation of AI tools and their high-profile abilities will encourage this trend, as well as their use to evaluate applications and to make recommendations about whether they should be accepted. The Rest of the World article points out that OpenAI, the company that is behind ChatGPT, updated its user policies in late March with the following as “Disallowed usage of our models”:

High risk government decision-making, including:

Governments trying to save money will doubtless use them anyway. It will be important for courts and others dealing with asylum claims to bear this in mind when there seem to be serious discrepancies in refugees’ applications. They may be all in the (machine’s) mind.

Follow me @glynmoody on Mastodon.

Filed Under: afghanistan, ai, asylum, chatbots, chatgpt, climate crisis, machine learning, openai, pashto, refugees, translation
Companies: openai

from the here-come-the-squeals dept

You may have noticed the world getting excited about the capabilities of ChatGPT, a text-based AI chat bot. Similarly, some are getting quite worked up over generative AI systems that can turn text prompts into images, including those mimicking the style of particular artists. But less remarked upon is the use of AI in the world of music. Music Business Worldwide has written two detailed news stories on the topic. The first comes from China:

Tencent Music Entertainment (TME) says that it has created and released over 1,000 tracks containing vocals created by AI tech that mimics the human voice.

And get this: one of these tracks has already surpassed 100 million streams.

Some of these songs use synthetic voices based on human singers, both dead and alive:

TME also confirmed today (November 15) that – in addition to “paying tribute” to the vocals of dead artists via the Lingyin Engine – it has also created “an AI singer lineup with the voices of trending [i.e currently active] stars such as Yang Chaoyue, among others”.

The copyright industry will doubtless have something to say about that. It is also unlikely to be delighted by the second Music Business Worldwide story about AI-generated music, this time in the Middle East and North Africa (MENA) market:

MENA-focused Spotify rival, Anghami, is now taking the concept to a whole other level – claiming that it will soon become the first platform to host over 200,000 songs generated by AI.

Anghami has partnered with a generative music platform called Mubert, which says it allows users to create “unique soundtracks” for various uses such as social media, presentations or films using one million samples from over 4,000 musicians.

…

According to Mohammed Ogaily, VP Product at Anghami, the service has already “generated over 170,000 songs, based on three sets of lyrics, three talents, and 2,000 tracks generated by AI”.

It’s striking that the undoubtedly interesting but theoretical possibilities of ChatGPT and generative AI art are dominating the headlines, while we hear relatively little about these AI-based music services that are already up and running, and hugely popular with listeners. It’s probably a result of the generally parochial nature of mainstream Western media, which often ignores the important developments happening elsewhere.

Follow me @glynmoody on Mastodon or Twitter.

Filed Under: chatgpt, china, copyright, generative ai, mena, mubert, music, openai, tencent
Companies: anghami, mubert, openai, spotify, tencent

DailyDirt: Open Source Artificial Intelligence Is Smart

from the urls-we-dig-up dept

The importance of machine learning is becoming clearer as vast amounts of valuable data accumulates and human minds are looking ill-equipped to try to parse all of it. Sure, humans seem to be better at deciphering our own handwriting and voices, but artificial intelligence (AI) might be a tad better at predicting long-term weather patterns or imminent economic instabilities. We’ve already seen AI that can beat the best of us at Jeopardy! and chess. Here are just a few open source projects that could help achieve the next milestones of AI and machine learning.

After you’ve finished checking out those links, take 10% off any $50+ order from our Daily Deals using the promo code DAILYDIRT.

Filed Under: ai, artificial intelligence, big sur hardware, darpa, deep learning, deepdive, elon musk, machine learning, oaqa, open compute, open source, openai, opencog, tensorflow, uima, watson
Companies: amazon, facebook, google, ibm, microsoft