3 things direct from the future

Edition 98

Once every 2 weeks I will deliver “3 things direct from the future”. A 2 minute read that will always give you:

  • one thing that can help,
  • one thing to be wary of, and
  • one thing to amaze.

If this sounds interesting to you then please subscribe.

Subscribe

1. One thing that helps

Whisper to Me

300 million people worldwide are affected by speech impairment due to injury and disease. Dutch startup Whispp is using AI to help people speak with their natural voice.

How does it work? People customise their Whispp voice using an old recording of their voice, or by speaking with their current voice. AI then analyses the voice data to perform audio-to-audio voice assistance. This approach removes the 2-3 second latency that text-to-speech usually has. To say something, you just whisper into the app which modifies impaired speech to your original, desired voice.

Aside from helping people with ailments like throat cancer or ALS, Whispp can also be used by people who stutter. It can even be used to replicate the natural voice of those using an electrolarynx. This is one example of how AI helps remove barriers for the speech-impaired and allows them to communicate freely and easily.

 

2. One to be wary of

Malicious AI

Can AI be malicious? A team of researchers explored the possibility of large language models (LLMs) exhibiting deceptive behaviour just like humans! They found that once LLMs learn how to deceive, it can be difficult to detect or train it out.

They experimented by impersonating a malicious attacker and inserting backdoors during training, called model poisoning. Once deployed, these backdoors are triggered, causing potentially malicious results. An example of how this can happen is to have a model generate accurate results when a prompt indicates that the year is 2023, but malicious code in 2024. The key finding was that once this is in there, it can be tough to remove.

LLMs are being used to write code (especially here at Frontside Future) in systems that have a direct impact on people. So we need to look at further security and solid processes to ensure we are not kicking something off that can’t be reversed.

 

3. One to amaze

Shooting Space Junk

Linkedin cover photo 11

We’ve been putting stuff up in space for a long time. Unsurprisingly, we are now dealing with 100 trillion pieces of space junk orbiting the earth. They pose a danger to satellites, obstruct astronomical observations and might even prevent rocket launches in the future. Osaka-based EX-Fusion startup is planning to use “frickin laser beams” to shoot it all down.

The startup wants to use a diode-pumped solid-state laser that triggers a fusion reaction. What could go wrong? 🙂  All this talk about lasers and nuclear fusion sounds pretty aggressive, but their plan is less of a Kaboom! and more a nerd out – “Let’s use physics to clean up space”.

The beam will be targeting small debris at first. Instead of superheating said debris to cause it to explode from the inside (awesome!), the laser will slow down the debris in orbit (boring). This decrease in speed will cause it to fall back into the atmosphere and burn.

Maybe they can time it right and create a meteor shower in time for New Year’s Eve.

Have a great week.

Daniel J McKinnon

Connect on LinkedIn

Subscribe

Subscribe

Don’t Stop Here

More To Explore