HN user
I read the title a couple of times and I'm still not sure it isn't misleading. The benchmarks are not just for Postgres but for Postgres with the Mooncake extension. There are also other results for Postgres with different extensions. While it does rank among the top fastest databases, it is not the fastest and not even within the top 10.
Do we know if Gemma models are fundamentally different from the ones hosted as Gemini? Gemini 1.5 flash seems to produce good results for the price and performance.
I was recently researching structured output generation for my project and I enjoyed using Outlines library a lot. It felt quite fast as it uses FSM and indexing. There are few fine prints though:
1. Sometimes constraints can decrease the quality of the output since syntax of the response is prioritized more than quality of the response 2. For memory constrained inferences, certain sampling strategies like top-k can cause OOM errors if the max_token is too high. I haven't tested that it is entirely due to structured generation but I suppose it is possible for certain regexes. 3. Vision models and other multi-modal models are not supported yet.
Apart from this, closed models also have json output but I am not sure how consistent they are
1. https://platform.openai.com/docs/guides/text-generation/json... 2. https://docs.anthropic.com/en/docs/build-with-claude/tool-us... 3. https://ai.google.dev/gemini-api/docs/api-overview#json
The author did enroll his kid into school for freshman after a break. Is this normally considered beneficial? Having a period of unschooling followed by traditional school?
Looks like some of the docs are generated by an llm. I see pictures with typos and imagined terms, incomplete texts etc., I wonder to what extent we can trust rest of the docs.
https://github.com/microsoft/Phi-3CookBook/blob/main/md/04.F...
On a tangential note, I find window management in MacOS much more horrible than Windows. Want to split windows, you end up with full screen. When on multiple monitors, selecting an app on one screen makes the same app active on the other screen (or sometimes it doesn't). I am willing to rewire my habits if I can just figure out how to make Mac window manager behave deterministically. I just don't get what is the grammar of user interaction that the designers went for.
This is really shoddy reporting. The title says Azure data breach whereas the attack is a phishing campaign targeting Office 365. Being a phishing campaign, it is unclear which components of Azure/Microsoft were instrumental in the attack. And the article goes on to make irrelevant allusions to Microsoft's negligent cybersecurity practices. This is such bad reporting that I wonder whether it is done with incompetence or malfeasance.
This is pretty impressive, it seems that OpenAI consistently delivers exceptional work, even when venturing into new domains. But looking into their technical paper, it is evident that they are benefiting from their own body of work done in the past and also the enormous resources available to them.
For instance, the generational leap in video generation capability of SORA may be possible because:
1. Instead of resizing, cropping, or trimming videos to a standard size, Sora trains on data at its native size. This preserves the original aspect ratios and improves composition and framing in the generated videos. This requires massive infrastructure. This is eerily similar to how GPT3 benefited from a blunt approach of throwing massive resources at a problem rather than extensively optimizing the architecture, dataset, or pre-training steps.
2. Sora leverages the re-captioning technique from DALL-E 3 by leveraging GPT to turn short user prompts into longer detailed captions that are sent to the video model. Although it remains unclear whether they employ GPT-4 or another internal model, it stands to reason that they have access to a superior captioning model compared to others.
This is not to say that inertia and resources are the only factors that is differentiating OpenAI, they may have access to much better talent pool but that is hard to gauge from the outside.
This model is built upon the Würstchen architecture. Here is a very good explanation of how this model works by one of its authors.
Of 1,463 proteins analysed, aided by with a type of artificial intelligence known as machine learning, 11 proteins were identified and combined as a protein panel, which the researchers have shown to be highly accurate at predicting future dementia.
I understand that press releases are intended for non-technical folks but I don't get the point of this description. Is it assumed that machine learning is less understood than artificial intelligence?
To what extent is this article impartial? Besides the inclusion of carefully selected quotes from executives, the remaining references predominantly originate from articles published on the same website and authored by a specific subset of writers. Is this viewpoint commonly shared within the industry?
It can also go the other way. You may just have to figure out patterns to jailbreak a handful of foundational models and profit forever.
I don't know why everyone is fixated on the term incognito. I believe what GP wants (and me too) is to pause watching history or to at least exclude a viewing session from the recommendation algorithm.
My problem with Youtube Premium is that it doesn't work if you are incognito. I don't understand why I need to sell my data if I am paying for a service. If I watch a cute animal video on a whim, I would get inundated with animal videos.
Correct me if I am wrong, to use LORA fine-tuned model in inference you would still need the original model + trained additional layers, right?
If we can perfect methods to fine-tune large models for specific task while reducing the overall model size, then it can fit into more consumer grade hardware for inference and can be broadly used. The objective is to prune unnecessary trivia and memorization artifacts from the model and leverage LLMs purely for interpreting natural language inputs.
I wonder how much these tropes and storytelling techniques shifts the baselines of cultural norms. For example, a sitcom would often find it convenient to have friends meet daily in a bar and somehow drinking almost daily is perceived to be a normal thing to do.
A simple metric on confidence interval could do the trick. As the model grows larger, it is getting more difficult to understand what is going on, but that doesn't mean that it needs to be a total black box. At least let it throw some proxy metrics. In due course, will learn to interpret those metrics and adjust our internal trust model.
actually if you hover over 3rd character, the visualization prunes the irrelevant path.
I just cancelled my Duo subscription yesterday. It doesn't make sense to pay for a terrible UX, lacking basic features that even free software has. I feel the apps haven't moved past MVP and they resolutely refuse to improve.
I have been wanting to cancel the subscription for months now but just got around to do it. I wonder what percentage of the current subscribers want to get rid of the subscription but simply couldn't bring themselves to do so.
Do you have any tips or guides for creating good training programs? I find it challenging to think from a newcomer point of view and I significantly underestimate their challenges. Consequently I feel the learning curve I set for the team is either too steep or too shallow.
This happens to me irrespective of whether I went through the painful learning curve just recently or several years in the past. Once I am comfortable with a topic, I cannot approach it from a newcomers perspective.
So I think having a systematic approach to KT and training will help.
Are there any Blinkist kind of service providing SRT prompts for books?
Is there a good summary of how AlphaFold has made a difference in the few years since it has been released. I was quite excited to hear their success in 2020, so I am curious to see where have we got to in 2 years and what is in the horizon.
I came across an interesting take on this in the book Algorithms to live by [1]. They argue that old age is a time of exploitation (as in explore/exploit strategy) where we tend to prioritize relations/experiences that we know to be rewarding. At a younger age, we are more accommodating to explore new experiences. That is why a teenager entering a dorm in a new environment full of strangers can get excited while for old people entering a retirement home it can be a harrowing experience.
It will be interesting to observe what this means for growth of cities. In the last 100 years, city sizes increased while staying within Marchetti's travel time budget [1] (1 hour for round trip) thanks to higher speeds available through motorized transport. Now that technology is redefining proximity once again, I wonder what factors will continue to encourage/discourage people to live in dense cities.
1 - http://www.cesaremarchetti.org/archive/scan/MARCHETTI-052.pd...
How about building design? I know in colder climates buildings are designed to retain heat. Are buildings in warmer climates designed with similar heat insulation in mind? I think traditional architectures are designed for convective cooling but that doesn't play well with air-conditioning. Not too mention the modern designs obsessed with glass which I think traps heat. Good and efficient for cold climates, not so much for the hot ones.
Teams is terrible to use for occasional use with personal login. I always had issues using it until I figured out that I need to empty the app cache before joining a meeting. Whenever I see a Teams link in an invite, I join 5 minutes ahead of time to clear my cache and test my audio and video.
But I am told that there is no such problem in Teams with enterprise SSO. Also I have other friends who are quite happy with it. Personally, I can empathize more with OP. MS Teams is a terrible software which seems out of place among the modern era clients.
See related
https://youtu.be/vSnCeJEka_s - The death of Agile - Allen Holub
In some ways, error messages are like crutches for the users. Some of the badly written error messages makes sense to the author of the code but not to the user of the code. For instance, this helpful reply on SO actually made me chuckle when I first read it, since the implication was that the error message makes sense when you understand what is going on, whereas a good error message is supposed to work the other way around
All I can say is it is incredibly useful, not just O’Reilly but they have got collections from a lot of other publishers. It is really useful as a reference or when you want to deep dive into something new. To top it, if you get to complete couple of books cover to cover, it is even better.
I can’t say whether it is worth the subscription fee. I have got my employer to cover for my subscription for the last 2 years, but not sure if I will be willing to pay it out of my pocket.
Does it take a certain non-technical aptitude to market oneself better? I think those who fear interviews may find this challenging as well.