Training AI requires consuming copious amounts of data. Consume bad data and the output will be wrong. In computer science this is called garbage in / garbage out. In some AI models once a bad piece of information is learned it becomes exceedingly difficult to purge it.
Tuesday Links
- Why are some countries rich and others poor?
- What if a drop of an infant’s blood can warn of the risks of incurable diseases later in life. Should we do the tests? Should the parents be told? Experts aren’t sure. (NYT)
- Some 47% of Oregonians who have pursued “death with dignity” listed fear of being a burden as a motivation. (WSJ)
- Avik Roy: Put Medicaid enrollees in the (Obamacare) exchange plans.
- California uses drones to detect and fine people who set off and launch 4th of July fireworks.
FDA to Release Some Letters to Drugmakers about Concerns
The FDA sends letters detailing their concerns to the firms applying for new drug approvals. In the past, these letters were never made public unless drugmakers chose to release them to the public. When the information was released, firms often only redacted portions of the letters, often concealing any FDA concerns. Critics have complained that decisions made secretly do not provide the public – or other drugmakers – with information we should have access to.
Price-fixing Via Priority Vouchers
The Food and Drug Administration (FDA) has announced the Commissioner’s National Priority Voucher (CNPV) Program. The idea is to have a fast-track review – as short as one to two months compared to 10 to 12 on the usual timeline – for drugs that address one or more national priorities, such as pandemic preparedness, domestic manufacturing, national security, or serious unmet medical needs.
Source: American Action Forum