640 karma · joined February 1, 2022
Although at the same time if no one notices and continues to enjoy the reading, then maybe no one is harmed?
In my gut I would be very sad if ChatGPT content started flooding in
If you were eyeglasses, it's worthwhile to pay an extra bit - usually on the order of $50-$100 - to get a blue light filter built directly into your glasses.
Further, my optometrist specifically recommended that I use slightly underpowered prescription. I'm at about 90% of what my glasses prescription could be but he tells me that this means my eyes will strain less. It has seemed to work for me since after a few years of having to increase prescription every 1.5 years or so, it's starting to slow down. Could be anecdotal or caused by aging but something to bring up with your own optometrist if you have one.
However, I migrated to react-router 6.4 and it has a very similar concept of embracing web technologies and partials can be defined as an outlet of another page where only the data for the partial and dom for the partial is loaded if sub sub-tree is navigated in. They say they took this idea from emberjs.
I felt the same way as the authors of unpoly and many in this thread looking at most UI frameworks until I stumbled on react router 6.4, and it very much enabled me, a primarily backend engineer, to be productive in the frontend. I highly recommend it to people who are trying htmx but want something that is more IBM "No one ever got fired for picking react"
There are many rules of thumb that took the last 5+ years to discover but are now quite standard. You are nit picking on fully connected, but if we add dropout, weight initialization, and adaptive learning rate to what they said, then we are fairly close to being able at least get a deep architecture to overfit a toy dataset and be off to the races for then applying it to a larger dataset.
Build your network - The next thing is that you may have experienced a set of problems in your personal life or work but that set of problems may not be intersecting the set of problems you are suited to solve. It's worthwhile to get in touch with other people and ideally understand their business problems in particular. If you can identify and validate a business problem then you are almost there since you now have a short list of potential future customers or consulting clients, what have you.
Cultivate your solution space - You can start right away at identifying a niche of solution space you are a candidate to solve and try to keep getting better at that space so that you can be on the lookout for intersection between your capabilities and an interesting problem you validate.
Another important note is to be able to discern what is a promising problem and also not what YC calls "tarpit" problems. Tarpit problems are usually in the consumer space and are problems that everyone has like "I can't find a good restaurant" but in fact many projects have come and gone but it's hard to realize because the skeletons of have sunk into the tarpit.
EDIT: to answer original question.
For me personally it had to do with maintaining networks of people I had worked with in the past in order to get exposed to business problems that I wasn't aware of.
In some projects I'm working on, I'm seeing close to 95F1 score using deep learning to do token classification based NER. However, using non-deep learning approaches (still training statistical models) I can get to 91F1 on my use case but have a much faster inference and not need to use a GPU.
I'm similarly optimistic about LLM, but fear that a lot of other really useful workhouse algorithms and strategies are going to be pushed aside and people will forget about them.
Back to my use case, some strategies I'm looking at are using cheap and fast CPU powered models/smaller models for inference and then based on certain signals decide whether a particular instance should be passed to a GPU based model for better accuracy.
Do you mean being in hacker news ?
For instance there is a project that I think is pretty interesting but it's built on top of kakfa so you pretty much need to be running kafka to use it. I wish I could swap out redis as pubsub + kv store and I'd use that project in a heartbeat.
I've heard of cases where vendor A gives a big discount to the tune of a the most recent year of annual billing from vendor B if you can prove that you most recent bill from vendor B is 0 due to completing the migration.
Even if you rely on streaming like PubSub or Kinesis, one thing teams I've worked on has done is to write interfaces in the application tier that allow for using an on-prem primitive like Kafka and not depending too much on the implementation of that abstraction.
I've been on a platform team that built these primitives into the application layer, e.g. a blob storage interface to access any blob store whether it's on prem NFS, azure, etc. However I'm looking at newer projects like dapr [1] and have taken them for a spin in small projects. Such a project seems like a favorable way to add "platform services" to a non-trivial app while still maintaining a pubsub abstraction that allows for swapping out the physical backend.
So agreeing with you but with the caveat that you can rely on platform service interfaces and then the service behind that interface could be a cloud vendor product or a boring technology provide you don't let that abstraction leak and call a very specific kinesis feature for example.
[1] https://dapr.io/
Another strategy for getting intro to users of OSS software is to join discord and start offering help and then possibly leveraging that to mention the help could be more permanent.
Do be careful though not to be explicitly soliciting since that is usually against the community policies so a little bit of a needle to thread.
I have my 7700x limited to 65w with very little performance loss -- it's very easy with eco mode. I've seen lot of benchmarks with 7000 series chips on eco mode but not so much on Intel side.
These chips really should have been power limited with a little less multicore performance instead of being power drains.
I was in a similar boat but on the analytics side wrt to maths. I also took courses at a local university in Math to build up my knowledge. Learning pure math at the college level, like learning to do proofs in front of people on a chalkboard -- I think is very difficult to do unless you take it in the context of a class.
I think you should still give it a go, it's never too late!
edit: answer original question
I was in data analytics at a health insurance company in more of a BI and data pulling role. Taking his course gave me the confidence to start doing more machine learning related projects in my company eventually becoming the first person in the company to be called "Data Scientist" after developing multiple models that were making an impact on the business.
edit: answer original question
https://www.coursera.org/specializations/machine-learning-in...
Statistical rethinking by Richard McElreath also helped me to understand bayesian analysis and simulation -- possibly the best hands on bayesian analysis book for beginners
https://xcelab.net/rm/statistical-rethinking/
edit: added links
The overhead to do a similar service using threadpools is much higher and the reason asyncio exists.