In yet another case of people’s privacy being invaded by the technology companies, it has been revealed that video clips recorded in homes through Amazon Cloud Cam are being viewed by humans in order to train the AI algorithms in responding to situations.
Though Amazon’s spokesperson has denied that private videos were being reviewed, there have been complaints received through people from within the programme reportedly being carried out in India and Romania. They may be called whistleblowers, in a season where whistleblower-related stories are hogging the headlines.
Amazon sells the Cloud Cam for $120 and the company provides free access to the recordings of the camera to the owners for 24 hours. This is typically useful for people who are traveling and would want to know if things are safe back at home etc. There are people who use the camera to monitor their children when being away using the Cloud Cam. If the customer pays a monthly subscription, Amazon extends this viewing duration to longer periods. Similar products in the market include Google’s Nest and Ring, which is also owned by Amazon.
Now if the gadget raises any false alarm, then the customer can send the clips to Amazon for it to check and do the changes necessary in the way the algorithms respond to certain images it clicks and sends the response back. Thus far, the situation is under control. What has now been reported is that the company’s teams are pulling out video clips for training the algorithms. In the normal course, the basic machine learning technology means the AI algorithms teach themselves. Using humans to train them is not something Amazon is supposedly doing. Again, the line is crossed when video clips recorded in the homes of unsuspecting customers are viewed without their explicit permission.
The issue had been raised earlier and the European regulators are very strict about privacy violations. There had been reports of some users switching off the Echo devices fearing their privacy is being invaded. A Bloomberg report on this claims Apple and Google had stopped using humans to train their AI algorithms. If this report is to be believed, Amazon is carrying on in India and Romania.
The real trouble is these AI-based security systems can send false alarms. People have reported that a cat jumping or a paper rustling have been identified as threats or raised as home-invader alarms. These can be annoying and the customer would want the gadgets to do a better job. Here’s where the fine-tuning of the algorithms becomes necessary.
The issue now has to be squarely addressed by Amazon and all its customers given a clear assurance that no video clips from the Cloud Cams will be viewed at their end by humans unless the customers submit for review and repair.