An open question is the degree to which governments can do good technical research, either in-house or commissioned. Some possible ways in which this could be relevant:

• Evals (although this likely could be contracted in a successful way) 
• More straightforward technical research (do any other safety-research areas use this? Nuclear security and areoplane safety seem to be done by the private sector and the role the of government is creating good incentives here) 
• The potential for governments to become the dominant actor in advancing towards TAI

Examples of governments (particularly US government) doing in-house or contracting technical research 

This list is not exhaustive–it’s just examples I happen to know about. 

Defense examples 

• Sandia national lab, the US governments lab for nuclear weapons safety. Run by Lockheed Martin since the 1980s. Hires an extremely large number of electrical engineers
• Lawrence Livermore, Los Alamos, and Oakridge national labs – develop US nuclear weapons 
• DARPA – US government acting as a grant maker for science research. Extremely good track record. 
• Cyber offence and defence work by intelligence agencies. My impression is that the US and the UK are 2 levels above private groups in cyberoffence and have a seemingly inexhaustible supply of zero-days. I’m unsure the degree to which this is done in house. 
• Generic defence contracting for new, high tech military hardware. 

Regulatory agencies 

• The EPA has a large in-house team of environmental scientists informing their regulatory policies. About half of their staff of the EPA are environmental scientists. 
• The Fed has an extremely large in-house team of economists who do economics research, often for regulatory purposes. The Fed’s economists are considered amongst the best in macro and financial economics. 

Other US government 

• US army corps of engineers. An unusual example since the military is such an unusual institution that does a very unusual about of in-house training with an unusual amount of market power over its technical staff. Also does engineering rather than technical research 
• NASA. Independent agency of the US government. 
• Department of Energy. Don’t know the specifics 
• Cybersecurity department within the department of homeland security. Are quite poor. 

UK government 

• NICE cost-effectiveness assessment of medical interventions for the NHS. Done in-house and my impression is that it is a very high-quality specific type of technical research which is statistics focused and reviews large numbers of existing medical research.
• UK independent regulatory bodies. Notably different from the rest of the UK civil service which, hire generalists, these organizations hire specialists, and these individuals stay in their areas rather than moving around regularly as is done in the rest of the civil service 
• The NHS does lots of biomedical research. Unsure how much of this is in-house. 

 Some patterns that emerge 

• Defence dominates government technical research 
• Reasonable balance of in-house and contracting. No obvious pattern in which is more successful.
• Some bias towards independent agencies doing more and better technical research. Particularly notable in the UK where there’s such a generalist focus in the civil service.