vivegi

No dispute on crediting Meta for LLaMa itself.

I dispute crediting Meta for the widespread interest in LLMs. IMO, that would be Google (for Transformers, BERT), OpenAI (for GPT-2, GPT-3). BERT and GPT-2 were released under Apache 2.0 and MIT licenses respectively and precede the release of LLaMa.

Of course, we can credit ChatGPT for bringing it to the forefront of public consciousness.

And none of this takes away from Meta's work on LLaMa. I had said as much in my original comment.

It is ironic that The New York Time Manual of Style and Usage [1] cited by the article as one of the authoritative sources often doesn't follow its own style guide. A Google search for "comprised of" in nytimes.com [2] gives around 4600+ results. A cursory glance at the first page of search results and the snippet text in the results all look fine.

I guess, we could say, much ado about nothing.

[1]: Siegal, Allan (1999). The New York Times Manual of Style and Usage. Three Rivers Press. p. 80.

[2]: Google search query for site:nytimes.com containing "comprised of". URL: https://www.google.com/search?q=%22comprised+of%22+site%3Any...

Except that the LLaMa model public release was involuntary (except for the research carveout); So, I am hesitant to give Meta credit for that while as they are also issuing takedowns.

Nevertheless, they are doing fundamental AI work. We can give them that.

There are around 32 cities [1] that will be hosting the events around the G20 summit in Sept 2023.

[1]: https://www.g20.org/en/g20-india-2023/host-cities/

Yes. It is said the Colonial British used to say The sun never sets in the British Empire.

In today's context with WhatsApp, The sun never rises without a billion GOOD MORNING messages and another billion GOOD MORNING TO YOU TOO replies on WhatsApp in India /s

... and here [1] is the original bet.

[1]: https://twitter.com/balajis/status/1638136293065658368?ref=b...

That is a can of worms. What is the operational definition of Rich?

  - Networth?
  - Income?
  - Net Taxable Income?

Instead, it can be an escalating penalty that is based on repeat offences; increasing fines for repeat offences, culiminating in something like losing driving privileges, impounding of the vehicle, permanent loss of driving privileges etc.,

I was wondering the same thing just a few days ago.

It is actually quite trivial to add an audio-stream (or streams) with the director/writer/actors commentary to HLS or MPEG-DASH and surface it in the application UIs.

I saw a very nice use of this in the IPL (India Premier League) cricket match yesterday. JioCinema is the streamer offering the match streams and they have an option to change the camera view. Of the multiple cameras in the stadium, you could pick a camera angle that you wanted. It was pretty cool.

For me, the sync is fine at the start of playback on PrimeVideo. It just becomes bad progressively (which leads me to believe they have used a video framerate that is ever so slightly different from the source and have keyframes insertion after a longer than optimal duration; similarly sample rate mismatch for output audio relative to input audio stream could be a potential cause).

And I use a FireStick, FWIW.

BTW, their own trascoder product MediaConvert seems to have this issue (It is possible that it could be user error too in how they have used the product or setup the parameters). [1]

My guess is PrimeVideo dogfoods MediaConvert and they also have this issue. They could have fixed it for newer content, but previously transcoded content still has issues (which will remain until they are re-transcoded).

[1]: https://repost.aws/questions/QUGajgu4zKTlewlTg1M96i_Q/questi...?

Customers, Business (i.e., Investors), People - of this trio of stakeholder categories, choose any two to satisfy.

There are relatively very brief periods in any company's lifecyle when they are able to choose all three and keep them all satisfied.

In a mature business, it is almost impossible to choose all three.

For big-tech, that era has ended and you see traditional CEO behavior making a comeback (i.e., regression to the mean). That is coming as a rude shock for many. That is probably what is happening with Google too.

Of all the video streaming services I have used, PrimeVideo is the one where the video/audio sync becomes terrible progressively.

It is pretty bad. It happens in 8 out of 10 movies. There is some misconfiguration in their AV transcoding pipeline.

And here, we have an article talking about Monolith vs. Microservices improving user experience.

Not cool. [1]

> What makes a cool URI?

> A cool URI is one which does not change.

[1]: https://www.w3.org/Provider/Style/URI

Meh. You don't get the list of every worker who was involved in the supply chain that ensured you got the laptop/tablet/phone/other device on which you are reading this. So, why is software or a developer special?

In this video [1] (NOT describing the paper), the professor is explaining Integrated Ammonia Plants and at 46:40 there is a slide that shows the Theoretical Minimum specific energy requirement at 20.9 GJ/ton Ammonia and the state of the art (since 1991) to be between 24-26 GJ/ton Ammonia (which is an extract from a textbook, Chemical Process Technology, John Wiley & Sons, Ltd. [2]).

How does the proposed method in the paper compare in its energy requirements?

[1]: https://youtu.be/-RZ0saG6Wq4?t=2800

[2]: https://www.wiley.com/en-kr/Chemical+Process+Technology%2C+2...

It would have been great if the KEYGEN element's semantics meant

  - generating a fresh keypair of a certain strength and persistence of the newly generated key if a keypair is not associated with a specific domain
  - selection of the keypair from the persistence store using a browser's UI dialog if a keypair is already associated with the specific domain, with an option to regenerate the keypair

That would have been a good declarative alternative for password-based-authentication.

It would have enabled client-side id generation and simplified user registration and other authentication flows, including perhaps keyrotation.

I guess we should use webauthn now.

It is there in the original CERN approval letter at the bottom of the article. CERN was relinquishing rights and putting into public domain the following three software artefacts and binaries.

  - W3 basic ("line-mode") client
  - W3 basic server
  - W3 library of common code

Absolute BS.

Correlation is not causation.

Its trivial. The obvious solution is to just simply read the extremely friendly documentation. /s

What is SM?

I guess the maxim Familiarity breeds contempt captures that.

If we look at the Top 10 oil producers in the world [1], India does not even appear in the list. If we look at the Top 10 oil consumers in the world [1], India consumes 4.79 million barrels/day or 5% of the share of the world consumption.

Having graduated to the most populous country in the world this year with a population of 1.43 billion people [2], energy security is a top strategic imperative for India. It has always been the case for independent India in the past 75 years.

From the article:

> Russian crude oil arrivals to India are expected to surpass 2 million barrels a day in April, representing almost 44% of the nation’s overall oil imports, according to Kpler data.

With that level of dependence on Russian oil for its energy security, it is highly unlikely for India to change its strategic calculus, Bloomberg article and First-world-pearl-clutching notwithstanding.

[1]: https://www.eia.gov/tools/faqs/faq.php?id=709&t=6

[2]: https://www.un.org/development/desa/dpad/publication/un-desa...

Software written in '98 on Oracle 8i/Sun/Solaris. Mostly SQL, PL/SQL & shell scripts. Still running even after db and os upgrades in 2023. It was mosly a database for online and batch transaction processing that fed a data warehouse and include the Extract-Transform-Load pipeline.

The reason for the longevity is the focus on extreme performance by focusing on I/O by utilizing the database and parallelism offered by the platform. Also, our coding language choices were pragmatic and are still supported. So, even with all the db and platform version upgrades, the code still runs.

You can (by introducing the dependency between previously independent events).

Its 2023 and still we have a zillion different ways

- of showing basic widgets like dropdowns, multi-select lists, toggle switches etc.

- of UI frameworks that require a lot of code and get obsolete at a rapid pace

- of dark patterns that are user un-friendly

If you take two mutually independent events and somehow introduce a dependence between the two, then the events are no longer mutually independent and therefore by law of conditional probability, you can influence the probability of the outcome of the combination.

Corollary: If you take two mutually dependent events and somehow break the dependence between the two, then the events are no longer dependent and therefore by law of conditional probability and law of independent events, you can influence the probability of the outcome of the combination.

A practical example of the corollary would be password re-entry user interfaces during account registration.

The reason while re-entering the password the second time, we are not shown what we typed earlier is to make the reentry independent of the previous entry. Otherwise, we may look at what was typed earlier and subconsciously type the same thing -- which would be bad while we are doing account registration.

Bithub is actually offering for Individuals

> $1.9/mo | $17.0/yr | $299.9/lifetime

(Not sure if they are A/B testing prices since the OP says $24/lifetime)

It is a no-brainer. I'll go with the monthly pay-as-you-go.

If after 6 months, I am still satisfied, I would upgrade to the yearly. If after 6 months, I am not satisfied, I would cancel.

The entire experiment would have cost me $12 worst case (and hopefully there were some leads generated and the cost paid for itself). That is just me though.

From OpenAI's Sam Altman https://playbook.samaltman.com/

Welcome to HN!

Try to write the documentation for the feature from a end user's perspective first. Stick to the bare essentials and effectiveness for the user.

Then try to write a brief about how the feature can be implemented using the developer's perspective. List down the translation units (i.e., source files), libraries/executables, databases/filesystem structures/files that will be impacted.

There will be multiple ways in which you could implement the feature. While writing this down, you can list all the alternatives briefly and then note down which alternative you have chosen and the rationale for it.

Then try to write a brief about how you will test it. At this time, it can be your overall test strategy and some key test cases (not necessarily all).

Keep all of this brief (you are not writing War and Peace).

With all this prep work, your mind would start using its associative capabilities to suggest how you can proceed with the actual coding.

A few comments:

1. If your idea is truly novel, i.e., you have invented something new, then you can patent it. But the question then becomes which geographies. For e.g., if you patent your work in Country X, then it is not protected in other countries. You will have to file in each country's patent office separately. If your idea is truly novel, then consider the costs of global filing of patents and evaluate against the benefits.

2. Software and algorithms are not patentable in many jurisdictions. e.g., India. Since you mention UI and backend implementation, it is very likely that it won't be patentable in many jursdictions that prohibit sofware/algorithm patents. For a more detailed analysis on software patentability conditions see: [1]

3. For sofware, you can take reasonable precautions against blatant copying by

- partitioning your software into server-side implementation that runs in hosts that you control and manage entirely

- client side software that requires run-time license keys (to control and keep track of usage, trial versions etc.,)

- focusing on security in your system from day 1 (includes source code security, release management, encryption, key management/distribution etc., some aspects of this fashionably called CI/CD or DevOps)

Having said all this, ideas are truly worthless. Iterating your product with paying customers is the hallmark of good execution. Execution is gold.

And remember: Most people will be thinking that they can implement it better than you and are unlikely to copy your stuff.

The hard part is acquiring customers, meeting their needs, addressing problems and retaining them by providing excellent service.

[1]: https://www.mondaq.com/india/trade-secrets/810286/is-softwar...

> blue skies and bs...

I am going to steal that line. :)