Our Analysts' Insights

Listen to This Blog
 

Artificial intelligence (AI) and caring people are together bringing life-changing services for accessibility and communication, but you must look past the hype and the deep fakes to find them.

Generative AI: Deep Fakes, Scams, and Misinformation

First, let’s get the problematic issues out of the way. New apps and services powered by artificial intelligence appear seemingly non-stop, (and some disappear, perhaps with your money, just as fast). Many are no doubt introduced by individuals wanting to make a quick fortune: Did you see how many subscription-based apps for ChatGPT appeared in your phone’s app store, even though the ChatGPT service itself is still (for now) actually free? While many added functionality and value to the ChatGPT UI, not all did, and some tried to imply – without saying so – that they were the official way to access the system. 

Many AI tools are also in the hands of those who want to influence our thoughts, mislead our beliefs, and encourage us to vote in a certain way. Some are blatant scams: deep-fake celebrity endorsements asking for your hard-earned cash. Where are the applications for the good of humanity, rather than someone’s bank balance or power complex?

And no, Morgan Freeman did not say this on TikTok.The manipulated image and cloned synthetic voice make it look and sound like a Morgan video, but it was not really the famous actor. AI engines and software that merges cloned voices – speaking many languages that the “source” person has never uttered – with manipulation and matching of facial movements are now able to produce videos that look so real it’s no surprise when viewers are fooled. Phone screens don’t emphasize digital tics like large screens do. It’s pretty scary when you consider all the misinformation AI can generate – either by accident, or by intent.

@voicechangertech Is sounds and looks SO REAL 😱 Deepfake + Morgan Freeman AI Voice #deepfake #morganfreeman #voiceai ♬ original sound - VoiceChangerTech

AI and the Greater Good

But it’s not all bad news. AI is already being used to change lives for the better. Some recent startups, new businesses, and university research programs are positioned to help people communicate by providing access to understanding in ways never possible before large data models and generative AI.

CSA Research interacts with language service providers every day: some focus on written word translation, others on interpreting. Many provide both services. But did you know that another form of interpreting – sign language – is often delivered by LSPs? It is certainly procured in the same way as spoken language interpreting by healthcare organizations, immigration, and legal teams. For many deaf people, sign language is their first language and therefore much more important than any written words. For those hard of hearing, captions and transcripts are often their preferred mode of understanding. 

There is a shortfall of qualified sign language interpreters – the US-based National Deaf Center recently reported serious gaps in provision of signers in education, the Register of Irish Sign Language Interpreters (RISLI) currently lists only 86 accredited interpreters, and according to the British Deaf Association, British Sign Language (BSL) is the preferred first language of more than 87,000 people in the UK – yet there are under 1,200 registered or in-training BSL interpreters. Worldwide, we see a serious gap in communication with little expectation of it being addressed any time soon.

So how can AI help? With virtual reality glasses, and with synthetic signers, through voice-to-text or text-to-sign applications! These two examples really shine:

• XRAI Glass pairs augmented reality glasses – the ones from Nreal that look like your favorite sunglasses rather than an alien’s headset – with a phone app and AI to produce live transcription of personal conversations, right before your eyes (Figure 1). While XRAI Glass does enable machine translation to nine languages today, the life-changing application is enabling deaf and hard of hearing people to better connect with others and understand conversations. 

Figure 1: XRAI Glass App - Conversation is Displayed in Glasses, Too

 
Source: XRAI Glass

• Signapse really made me smile – this startup takes the same type of AI techniques used in the deep-fake Morgan Freeman video linked to above, but instead builds something to genuinely help people every day. Signapse delivers text-to-sign: virtual sign language interpreters that act, look, and move like humans. Already assisting passengers on the UK rail network with British Sign Language (BSL) on train station platforms, expect to see Signapse expanding into additional languages and media soon. After all, there are at least 300 sign languages around the world – and many platforms for delivering information.

Figure 2: Announcements on Station Platform with Signapse Virtual Sign Language Interpreter

Source: TransPennine Express

AI is bringing great changes to all we do, from transcription and machine translation to synthetic voice and auto-generated images. Although human interpreting might be preferable, the reality is that it is not always available when people need it. These technologies can make a real difference in helping individuals participate fully in society and opportunities that they might not otherwise have access to. 

No doubt the world will see a myriad of new applications over the next few years – including more of those that will genuinely improve human communication for every individual in the global community. Look for those that embrace ethical practices and include features such as digital watermarks to prove origins, ensure traceability, and indicate what is and is not synthetic.