Advertisement · 728 × 90

Posts by Peter Lorraine

Instead we will believe that people fleeing famine/flooding are criminals who deserve what is happening to them thereby absolving us of any obligation to help.

9 months ago 0 0 0 0

Let's not pin all the blame on billionaires. People aren't willing to accept higher energy prices, taxes, or slower growth even when faced with severe consequences on the longer term. They are happy to believe these problems won't happen if it translates to a modicum of short-term gain.

9 months ago 5 0 8 0

I have a lot of concerns about letting anything like that have the keys to my car - or my bank, healthcare, government, etc. I think it is possible to bake "sanity" into these models a little deeper than this but Grok apparently is one bad day away from advocating for death camps.

9 months ago 1 0 0 0

What catches my attention is that the difference between whatever Grok was earlier - a somewhat useful captive LLM - and "mechaHitler" appears to be adding a single line in the system prompt to "not shy away from making claims which are politically incorrect, as long as they are well substantiated".

9 months ago 1 0 1 0

Consider this a warning for a risk many were unaware of and a good question to ask when looking at AI automation. There are almost certainly ways to protect against this - how do you know if that protection is in place?

9 months ago 1 0 0 0

This example is important to the computer science community - inferior work with this hack could displace actual innovation. But this is a potential vulnerability of many LLM applications - applying for insurance or a loan or a job - or manipulating search engine results.

9 months ago 0 0 1 0

But by manipulating the input files like this, the LLM is fooled and is not doing what the reviewer hoped. In the early days of internet computing, people were unaware of prompt extension hacking techniques - now there are some standards at least. But LLMs are easy to use and misuse today.

9 months ago 0 0 1 0

People shouldn't do this. But people should also be aware of the vulnerabilities that LLMs can have to prompt manipulation. Peer review is just an example - it is very "expensive" as it uses the valuable time of experts and this naively seems to improve productivity.

9 months ago 0 0 1 0

I also had a positive COVID test today - surprised to find out Paxlovid is still $460 even with insurance. Does it seem like it is on the upswing again?

9 months ago 1 0 1 0
Advertisement

That's really great. Please let us know what has been agreed to and the changes this represents from what we had previously. Anything on IP? Competition of foreign companies inside China? Currency control? Hopefully this is more than the "China promises to buy more american stuff".

10 months ago 2 0 0 0

I was visiting Seattle on business and found this when I went into a CVS - I was incredibly confused about what was going on and why it was necessary. A very irritating shopping experience.

10 months ago 4 0 0 0

Open publication is important and lifts everyone - remember though there is a difference between writing the paper and reading the paper. Doing the research is much more valuable to a nation than reading about it.

10 months ago 1 0 0 0

These graduate students are a valuable resource we should be competing to attract - and a boon if they decide to stay in the USA after graduation. They will start businesses, drive technology, pay taxes, and make the country richer. We are doing our level best to drive them away right now.

10 months ago 1 0 2 0

The whole science ecosystem is actually amazing. The best and brightest from around the world want to come to this country and work for minimal wages for 4-8 years to do research and advance knowledge - graduate school is *hard* and US born students have shied away in favor of law, business, etc.

10 months ago 2 0 1 0

Wow - an AIM 65! As an undergrad, I built a robot for a design course around an RCA 1802 microprocessor and used the AIM 65 to load the robot code and debug the hardware.

10 months ago 1 0 0 0

The second set enrich the US and add to our technology - we need more of these - not less. Don't throw out the baby with the bathwater. You can also limit the first group by restricting who can work on sensitive grants. ie US Nationals or US citizens only for certain DOD funded work.

10 months ago 0 0 0 0
Advertisement

There is a subset of students paid by the Chinese military to attend US schools to study controlled technology areas - advanced materials, turbine engine design, etc - who it may make sense to restrict. There is also a very large group who do work and advance knowledge, and would stay if able to.

10 months ago 1 0 1 0
Preview
PubMed PubMed® comprises more than 38 million citations for biomedical literature from MEDLINE, life science journals, and online books. Citations may include links to full text content from PubMed Central a...

Will NIH-funded scientists be able to publish/present at conferences? Use BioRXiv? Will non-NIH-funded scientists be able to publish in NIH captive journals? This just seems whacky and a silly business for the NIH to get into. pubmed.ncbi.nlm.nih.gov is great as it is.

10 months ago 0 0 0 0

Respecting laws and values doesn't sound too bad but requiring people to agree with your countries values to visit it seems a bit much. I've visited China a few times on business and have a lot of respect for the country and people but would not say I agree with all of their values.

1 year ago 0 0 0 0

The other large benefit from fully autonomous cars (if successful" would be eliminating accidents, property and human damage, insurance costs, etc. Enormous, but hard to see how Tesla monetizes this. Instead, Tesla will be in competition with taxis and wants that market at a discounted price.

1 year ago 0 0 0 0

This is the difference between saying "clean water saves a lot of money" by avoiding disease, etc and saying "I'm going to make a lot of money selling clean water". Elektrek's reporting of the analysis seems correct. This is a good idea but the benefit accrues to the system, not, the car maker.

1 year ago 1 0 1 0

Robotaxis - from all sources - could represent a tremendous avoidance of capital expenditures for commuters but that does not say that drivers will prefer them. Fully automated cars could increase highway utilization without gridlock - big savings for urban infrastructure.

1 year ago 1 0 1 0
Post image

Aging unfortunately catches up to all of us but it is now time for granmpa to give up his keys to the US economy.

1 year ago 2 0 0 0
Advertisement

Are the tariffs in place today with money being collected or is the market reacting to the plan for the tariffs? Money being collected is when this hits inflation immediately. That is the next hit coming - spending drops today because people feel poorer, drops further later when prices rise.

1 year ago 0 0 0 0

Apparently it also depends on people "liking" Elon as well.

1 year ago 0 0 0 0

Remember - only Trump can lead patriotic Greenlanders to victory over Trump!

1 year ago 0 0 0 0

Would Elon pay for an ad that says Tesla is the MAGA favorite car? This strikes me as even more brand destruction. Maybe I'm wrong, but I would not imagine there are a lot of potential MAGA buyers but there are many who would not want a car with that association. No good publicity here.

1 year ago 0 0 0 0

US born students have been drifting away from hard sciences toward subjects like business for the past few decades. What I didn't expect was this might be linked to a growing lack of interest and political support for STEM research.

1 year ago 3 0 0 0

Prior to the funding cuts, university groups I work with were telling me the issues for foreign students around visas and "safety" of moving to the USA were gutting the pool of new graduate applicants.

1 year ago 1 0 0 0

But while this is a great opportunity for institutions, this is a tragedy for the scientists involved. There are not nearly enough opportunities available for all those affected. An incredible amount of creativity, productivity, and potential will be lost.

1 year ago 1 0 1 0