The author makes a good point, but also, the visual appeal of this blog is very readable and appealing. Not a format I would have thought of but it's a pleasant surprise.
60 karma · joined March 1, 2023
If consumer hardware is already capable (in many settings) of reproducing what were formerly research-level and industry-grade techniques, it may be a transformation in more areas of technology than would be obvious. I am very curious to see if there will be further findings in this area.
IMHO a few factors need to be met:
* Truthfulness -- granted that many topics have opinions and schisms.
* Reliability (self-healing articles?)
* Ease of use -- I think this needs the most improvement. Search is iffy and navigation unfriendly, IMO.
If nothing else I think the growth of something like a competitor is beneficial on its own. A single "source of truth" is nice, but that's not the age we live in, for better or worse.
2. This is not strictly related to the article content, but I hope I'm not the only one disturbed by the low quality of writing coming from even AP. I don't try to look for nits to pick but this article is a good example. E.g. "triumph tinged with perfectionism." -- This is poor wording. I think I understand the meaning -- that perfectionism, which has downsides, has removed something from what is otherwise triumphant. But it is not written clearly. Another: "Nancy Cartwright arrived at her 1987 audition expecting to read for Lisa Simpson. She had other ideas." -- This reads like a line AI wrote. There are other examples scattered throughout the content.
I guess it's not really important, and I guess there's no reason for me to be picking on this article. But this is a top-of-the-line publication (in theory) and a relatively high-visibility article. I know writers are under pressure to produce content. But there are plenty of writers who perform well under pressure, and editors exist for a reason -- what does it imply that AP, among others, is disinterested in the quality of their own articles?
Maybe I'm reading too much into it. More and more patterns seem hostile, antagonistic to the user, and it seems like it's an adopted practice that's taken as a standard. I hope I'm wrong.
I think OSS is OSS always...being able to audit it makes it (reasonably) reliable, at least in the sense of security. I can look at the code, run checks, etc. That alone doesn't guarantee things can't crash and burn, but it's a great start compared to a closed-source solution, even if that solution stands on its promises, as reputation in software is an iffy prospect today.
Does anyone know when (if?) the offending passages got included in the published version?
What methods have you found to brute-force through the problem?
(The challenge of course is when you can't or aren't allowed to ignore it, its own challenge).
Like any AI tool, the created results can be positive or negative. The churn of poorly written articles and reduplicated imagery has most of us on edge, but that should not negate the purpose and quality of other output. Some commenters appear to see this tool as just another means to rapidly generate slop content, but I don't see why it should have to be, nor do I think that was the intent of the author / creator.
If what guides the use of the tool is the desire to create a certain piece of content, to express something, and the tool aids in telling the human story, then it's a winning answer, I think.
To my mind currently some tech jobs just get caught up in the whirlwind of self-justification, while others you could argue provide real value to people. Which side of that a particular role is on is, I think, largely subjective.
So...someone thinks it's worth it to pay you. Maybe you care too; at the least, it puts food on the table because it matters to someone enough to swap money for.
That's my 2 cents, but I invite a disagreeing perspective.
(As a side note which may be a kind of artifact of coding turf wars, I just can't consider a Node application with a number of dependencies truly lightweight, no matter how straightforward in practice. I know this is more lightweight than most comparably scoped applications, and the term lightweight often just implies fast / efficient / easy, but the overhead required for a standard Node JS setup appears contrary to the concept of lightweight in principal. Have at me).
It's inspiring I think that the depth of passion in any topic can carry you through to important discoveries. In that sense the coding hobbyist and the citizen scientist are little different IMHO.
The processing seemed to jump through a huge amount of video really fast...was it pre-processed or does it just rack up enormous amounts of compute (if you don't mind the question)?
Speaking of compute, is there a cost/fee expected to drop for this?
Thanks for making & sharing this tool.
IMHO it's premature to have a strong conclusion on this, and way too easy to let feelings, pro or con, influence our projections.
In the abstract the authors write "Among the existing studies, there is also divergent understandings." It'd be interesting to see the methodology of this paper (doesn't seem to have access without pay) and how it differs from previous studies.
That's just my take, again I don't think it's that bad. The article would be a useful breakdown for beginners.
(Also, I'm sure you know, LLM content sounds that way because the LLM was trained on content just like this, so it's not really surprising that a guide generated by an LLM would sound like the kind of guide that was used to train an LLM...)
It may end up moot -- the article reads a bit on the speculative side, and this project might fall the way of many startups of "good idea, bad implementation," but I'm willing to eat my words in the future on this.