1. One thing that helps
Artificial intelligence can find 20% more cancers in imaging scans than expert radiologists, one study has found. If true, that could be the difference between someone’s life and death.
Two groups analyzed 80,000 mammograms of Swedish women. The first group consisted of a radiologist working with AI, while the second group consisted of two radiologists, as currently is the standard. It was found that the radiologist-AI pair detected cancer in 6 out of 1,000 scans, while the radiologist pair managed 5 out of 1,000. Similar systems like MIT’s MIRAI have also been shown to increase accuracy in predicting a patient’s risk of developing breast cancer.
Does this mean that radiologists will ditch much of their work? Not necessarily. Dr. Laura Heacock, a breast radiologist from NYU says, “These tools work best when paired with highly-trained radiologists who make the final call on your mammogram. Think of it as a tool like a stethoscope for a cardiologist.”
2. One to be wary of
Can an algorithm predict your chance of being a victim of domestic abuse? In Spain they think so, but the system has some deep flaws.
Spanish police are using a system called VioGén to assess the risks faced by victims of domestic violence. It’s been under fire after women received low-risk scores only to be victims of homicide after their visit to the police station.
VioGén makes the assessment based on a 35-item questionnaire that the victim has to answer. The algorithm then assesses whether the victim has negligible risk, low risk, medium risk, high risk or extreme risk. This score determines police action – from follow-up calls to 24-hour surveillance. While Spain’s Interior Ministry claims the system is effective in curbing domestic violence, data since 2007 shows that VioGén had scored 247 of 990 women killed.
What went wrong? The police are relying too much on VioGén’s scoring. This is another case of offloading responsibility to technology instead of using it as another tool in the tool belt. Surely, just the fact that a woman goes to the police indicates something is wrong.
3. One to amaze
Therapists often suggest writing a letter to your past or future self as a way to resolve deep-seated issues or to clarify goals for the future. But what if you could actually chat with “future you”?
MIT Media Labs have developed Future You to help young people connect with their future selves at 60 years old. They’ve used AI to age your present self to how you would look at 60, just like those filters you see on social media. Then, using GPT 3.5, the researchers created synthetic memories based on the participants’ information, anecdotes and their ideal future selves.
After chatting with their future selves, the participants, aged 18 to 30, exhibited lower anxiety and felt more connected to them. The researchers hope that this approach will promote better decisions and behaviour for the participants, such as living a healthier lifestyle or putting more effort into their careers and studies.
I know 48-year-old Daniel could definitely help 19-year-old Daniel. I’m just not sure he would listen.