HN user
It's not addictive along the same pathways (as far as I'm aware) from what I observe in my own life you either smoke casually and could stop whenever you want to (I used to then I stopped no problem) or you are dealing with psychological stress factors that are best treated with the help of a mental health professional.
The study controlled for general health but I would note that the odds-ratio reported for cannabis use (4-6) is not so different from the odds-ratio associated when studying high vs. low income cohorts (3-5). This study is not a random blinded clinical trial it's reporting a trend present in ~5 million electronic health records which I assume record yes/no to patients ever telling a doctor they have used cannabis.
I would assume that cannabis use correlates with a few other important heart health variables and we would expect the odds ratio to be lower when accounting for those (alcohol doesn't have an OR more than 1.0, tobacco smoking ~1.5)
I'm sure that cannabis use is bad for cardio health but the reported odds ratio is very high. I personally do not use cannabis.
I'll start my first job in an engineering role in a few days and I could reach behind to my bookshelf and flip through "The Way Things Work" right now if I wanted to. Fantastic book that really inspired me when I was younger.
I think the 600 dollar price is more than double the price of the same diplay as a mass-produced product it's a price for enthusiasts of the technology
and it's open source so nothing stops a bigger producer of copying the exact technology with institutional funding and manufacturing expertise
that is almost guaranteed an at-cost production figure for the limited run of kickstarter funded displays there isn't a production line producing these things - watch the youtube video this guy quit his job for over a year to build a passion project into a prototype
This is a linear regression relying on a couple years of data to predict 15 years in the future and I don't believe that the valuation is made on this basis.
It may be that spaceX is buying an operation that would realistically take 5 months and 100 million to copy in-house for 60B because the worry is that waiting 5 months might cost that much in some sort of lost opportunity. It also might be that in any negotiation SpaceX is viewed as incredibly cash-rich and so anything can be sold to them for inflated prices.
I really don't understand these companies valuations it seems like boardrooms everywhere are in a constant state of panic that they'll lose it all if they aren't growing a breakneck pace constantly.
I've been wondering lately if it would help to take a medium sized model and either in cloud or some local setup actually do Reinforcement Learning from Human Feedback (RLHF) on every prompt as a chore - I don't know if trying to manually finetune a model to your use habits would ruin it or help - ideally if you were diligent you could get rid of some of the ticks that make models for the general public difficult to work with e.g. overly sycophantic, overly verbose, annoying tendency to explain via analogies
but perhaps one individuals prompt feedback just isn't going to ever be enough I'm not sure how much you need (I know people working at big companies that have purchased in-house agents fine-tuned on internal documents etc.. and apparently these end up with bizarre behaviours not necessarily more helpful than the standard models)
I'd like to be able to essentially edit every response given by an agent and then finetune on the difference between what it produced and how I edited the text. Personally I would just remove a lot of the adjectives and try to distill the responses to core responses but I worry based on some of the work done by Owain Evans and other alignment researchers that this can sometimes push agents into tricky-to-predict tendancies.
exactly yeah it was a code base written by atmospheric physicists I assume and I had an idea that maybe copilot could get it working to interface with some more modern software and it just didn't really have what it takes.
Even with 3 weeks I'm just not the Fortran/C programmer to get that job done so I moved on to other things.
you assume correct
It's funny lately I've been seeing the cursor advertisements all with some premise of regular young person wants to develop an app and the ads really do focus on the simplest of premises: the only ones I've seen in these skits are essentially variants on the "todo app" web app tutorial
the tech is pretty good at helping identify simple bugs when they happen and to write short sections of code given very explicit instructions but yeah I have yet to see good examples of short one sentence ideas turned into a working product that looks better than anything that could be a UDemy tutorial app.
On the post-grad job hunt right now - I note that most employers will ask in a technical interview or whiteboard interview "how are you using LLMs?"
It's tough to answer because you want to hedge for both an AI enthused employer and an AI hesitant employer with limited information about who they are and how they personally use these products. I've been responding with a sort of long winded answer about how 'there is clearly a learning curve for how this technology fits into any process and how I always always always double double double check yadayadayada'
I'm probably using the chat/ask functionality on a daily basis for quick debugging / new technology learning questions but I have yet to really use the fully agent or computer-use products because I've had more bad results than good the few times I've tried them (re-factoring a big repo of decades old fortran+C code for modern compiler/OS some things started to work but ultimately I abandoned that effort).
Love that guy - you've truly lived if you can get a line like this on your wiki: "Given Bull's past ventures, it has been speculated that besides Iran or Israel, the CIA; MI6; or the Chilean, Syrian, Iraqi, or South African government could have been behind his assassination."
its funny that sky features like clouds and blue patches end up being fit very close to ground level because there isn't a difference in perspective to cue in the algorithm that the skybox should be tall, I wonder if there is any way to incorporate the fitting algorithm with simultaneous lidar data about true distances of things to allow scenes to be viewed from further up
Wishful thinking but it would be real great if an engineer poisoned these datasets with bait entries
It may be simpler to build from scratch using parts from a hobby store if you want a drone which cannot be tracked back to you or your credit card
If a cloud vendor with 1 million users experiences a long term outage: the vendor has a serious problem. If a cloud vendor with 1 billion users experiences a long term outage: the internet has a serious problem. Yada-yada-yada xkcd/2347 but it's the big block in the middle which crumbled
The number was correct to a reasonable degree under the assumptions stated by the author in the paper that tweet references since they obtained estimates from consumer grade hardware and the carbon intensity associated with average kilowatt produced in the United States not a hyperscale datacentre run using "ML best practices" although this distinction is left out of various lay media citations. The number also did not pertain to inference it was associated with training a particular model from pre-2019.
Agree on training. But that google paper was written when the only image model available for broad public consumption was dall-e 2 and video models were more than a year away. It gets a mention in a more recent 2024 paper [1] which goes into detail about how inference rather than training creates the difficult to manage energy load which grids struggle to meet. If consumer interests and demands drive the trend in what companies offer in terms of inference capability then it's fair to worry that the impact on sustainability goals will be an afterthought.
Yes but the task becomes that much harder - we are scaling up natural gas generation not to phase out coal but simply to meet demand that wouldn't exist without the fierce competition to build the biggest LLM. Any feasible plan made 5 years ago which may have worked to transition a large industry from fuel burning energy sources to electricity generation (renewable or otherwise) is made 10x harder by the introduction of this rapid rollout in datacentre capacity.
This is a common agreement to have with industrial power users. I know in Quebec during the coldest days in winter industrial users are required to scale back.
I would hope there aren't too many large utility jurisdictions which would curtail citizen consumers in favour of industrial users in the event of a demand surge.
On a related note. It's worrying to me how quickly we've accepted that we're going to boost electricity consumption massively prior to achieving anything close to the carbon intensity reduction targets which would mitigate the worst of climate change effects. It's all driven by a market force which cannot be effectively regulated on a global scale for multinational tech firms who can shop around for the next data centre location with near total freedom. And with advances in over the top fibre networks etc... a tonne of AI demand can be met by a compute cluster on the other side of the world (especially during model training) so the externalities related to the computing infrastructure can theoretically be completely dumped somewhere far away from the paying customer.
There is more than one comment here asserting that the authors should have done a parallel comparison study against humans on the same question bank as if the study authors had set out to investigate whether humans or LLMs reason better in this situation.
The authors do include the claim that humans would immediately disregard this information and maybe some would and some wouldn't that could be debated and seemingly is being debated in this thread - but I think the thrust of the conclusion is the following:
"This work underscores the need for more robust defense mechanisms against adversarial perturbations, particularly, for models deployed in critical applications such as finance, law, and healthcare."
We need to move past the humans vs ai discourse it's getting tired. This is a paper about a pitfall LLMs currently have and should be addressed with further research if they are going to be mass deployed in society.
Reading this reminded me of this good lecture on the physics of racing which was posted to youtube.
Andre Marziali - Physics of Racing https://www.youtube.com/watch?v=bYp2vvUgEqE
-Naïve- is the example I think I see most often but I think it’s often spelled -naive- and no one would fuss too much with either spelling
Ah yes, I turned off a bunch of Kaspersky internet security settings and I'm through. This is my work computer I forget what's running in the background sometimes.
In both Edge and Firefox I'm blocked for using Adblock but from what I can tell I do not have adblock on either browser.
This has been on my mind for awhile and is why I only briefly used copilot on a daily basis.
I'm at the beginning of my career and learning every day - I could do my job faster with an LLM assistant but I would lose out on an opportunity to acquire skills. I don't buy the argument that low-level critical thinking skills are obsolete and high level conceptual planning is all that anyone will need 10 years from now.
On a more sentimental level I personally feel that there is meaning in knowing things and knowing how to do things and I'm proud of what I know and what I know how to do.
Using LLM's doesn't look particularly hard and if I need to use one in the future I'll just pick whichever one is supposedly the newest and best but for now I'm content to toil away on my own.
"British money is in pounds, shillings, and pence. The British are used to this system and they like it, and all your arguments that the American decimal system is better won't convince them."
Give them time...
It continues to seem like this administration sets out to achieve 200% of each objective so that they can retreat to a position where they have attained 90 or even 100% of their original plan after a series of court decisions and popular backlash.
Every institution which could in theory check executive overreach is either already captured by ideological allies or is completely overwhelmed.