Suramya's Blog : Welcome to my crazy life…

May 18, 2023

Lost in Translation: Movie Synopsis edition

Filed under: Humor — Suramya @ 10:52 PM

This screenshot of the synopsis for ‘Alien’ popped up in my feed earlier and I couldn’t stop laughing.

“Space ship people get up from sleeping coffin and have eat. Computer woman find strange noisings on planet and astronauts go to seeing. Astronauts find big elephant man who dead then find too many egg/

Astronaut is possess by egg demon and new egg demon is come when eat bad noodle. Seven friends and cat all try to find egg demon before ship go home but is hard working.

Who will life to escaping? Who is bad milk blood robot? Scream not working because space make deaf.”

This reminds me of “All your base are belong to us” which was another masterpiece of bad translation.

– Suramya

May 17, 2023

Request to advertise ‘Men’s Beauty’ topics on Linuxgazette.net

Filed under: Humor,Linux/Unix Related — Suramya @ 7:18 PM

As some of you might know, I host a mirror for Linuxgazette.net and this usually results in an email every couple of months for changes to the existing articles and requests to advertise. I recently received an email requesting information about linuxgazette.net for advertising. Now there are a few problems with this, firstly I don’t host linuxgazette.net I just host a mirror of the site. Second problem was the content/topic they want to advertise on the site:

Beards seem to have become popular again, but do you think they will stick? Time will tell, but personally, I wouldn’t mind seeing more men with a cleanly shaven face 😍

I’m getting in touch today because I have some clients looking for advertising opportunities that are related to Men’s Beauty (Should I call it Men’s “Beauty”). After I stumbled across linuxgazette.net I figured I’d reach out and see if you ever work with brands, or have a media kit available.

I mean historically Linux admins do have a reputation of keeping big beards but this is ridiculous. Of course I know that this mail was probably sent out by an automated bot but it did make me laugh .

– Suramya

May 16, 2023

Using AI powered chatbots to answer questions about a book series

One of the things that I have been wanting for a while is the ability to look up stuff about a book or series so that when a new book in the series comes out I don’t have to re-read the old books before starting the new one. I end up doing that because otherwise since I have forgotten a large part of the backstory it would make the new book a lot less enjoyable. There are some authors who solve this problem by summarizing the backstory when the book starts and it is really useful. When ChatGPT and associated AI bots started picking up one of the ideas I came up with was to create a chatbot trained on a book series so that I can ask it a question about the book and get an answer. Would be really useful when dealing with book series with a large cast and multiple overlying stories.

Recently privateGPT showed up in my feed with the following claim: “Ask questions to your documents without an internet connection, using the power of LLMs. 100% private, no data leaves your execution environment at any point. You can ingest documents and ask questions without an internet connection!” and it seemed like it was the perfect tool for what I wanted to do so I downloaded the program from GitHub and set it up on my machine.

Once the software was installed, I copied a single book from the Mercedes Lackey Valdemar series called Darian’s Tale 01: Owlflight into a temp directory and told privateGPT to ingest the contents. The ingestion took hours even though the file was only ~600kb in size, I left it running overnight and it completed sometime during the night. Then I ran the prompt to be able to query the chatbot. The first few times I tried querying it, the app crashed because I ran out of memory. So I enabled SWAP on my system and then it finally ran the query and answered.

Unfortunately the answers I got were complete nonsense. You can see the questions I asked and the answers I got below:

Enter a query: what is a heartstone

llama_print_timings:        load time =   974.93 ms
llama_print_timings:      sample time =     0.00 ms /     1 runs   (    0.00 ms per run)
llama_print_timings: prompt eval time =   974.77 ms /     7 tokens (  139.25 ms per token)
llama_print_timings:        eval time =     0.00 ms /     1 runs   (    0.00 ms per run)
llama_print_timings:       total time =   977.55 ms

A heartstone is a type of music that is typically composed in the key of F major. It is characterized by a steady rhythm and a melody that is often slow and mournful. 
Heartstones are often used in classical music and are often associated with love and loss. They are also used in some contemporary music genres, such as folk and world music.

The correct answer is (as per Valdemar Fandom Wiki):

A Heartstone is both an artificial node produced by the Tayledras and the node’s physical anchor. The knowledge of how to create a Heartstone was given to the Tayledras by the Star-Eyed after the Sundering of the Clans. She tasked the Tayledras to cleanse the untamed magic caused by the Mage Wars. A Heartstone is used as a vessel to contain the cleansed magic. Ley-Lines are re-routed by Adepts and Masters to drain entirely into the Heartstone, bringing the surrounding area back to normal. Once their work is finished at that location, the Clan will establish a new Vale and Heartstone in a new area that needs cleansing. They drain the old Heartstone entirely and transfer the energy to the new Heartstone where they will begin their work anew.

So I asked another question:

Enter a query: who are the Hertasi

llama_print_timings:        load time =   974.93 ms
llama_print_timings:      sample time =     0.00 ms /     1 runs   (    0.00 ms per run)
llama_print_timings: prompt eval time = 26948.27 ms /     7 tokens ( 3849.75 ms per token)
llama_print_timings:        eval time =     0.00 ms /     1 runs   (    0.00 ms per run)
llama_print_timings:       total time = 26976.13 ms

The Herta si are a type of ghostly beings in the world of Eri. They are often associated with love and romance, and they are known for their ability to control and manipulate emotions. 
They are often depicted as beautiful, ethereal creatures, with long, flowing hair and eyes that glow with an otherworldly light. 

In this result it is talking about characters from a completely different series called the Herta. The correct answer is:

Hertasi are popularly considered to be one of the five non-human sentient races created by Urtho, but the work to “uplift” the hertasi was begun earlier, by Sorcerer-Adept Khal Herta. When he died, his work passed to Urtho.

The wild hertasi began as large, solitary lizards living in swamps, with the intelligence of a dog. They were “mild-tempered reptiles, available in large quantities, living fairly simple lives.”[1] They could have been described as semi-sentient before the Adepts took a hand. After Khal Herta’s efforts, Hertasi “had simple structures, organized hunting and fishing, and rudimentary medicine.”[1] Today, they are extremely active, intelligent, cave and tunnel-dwelling, social creatures.

Since the results were completely useless like all the other AI chatbots that I have tried so far, I have removed the software on my machine and will come back to this at somepoint later in time by when things would have hopefully improved.

– Suramya

May 15, 2023

Quickly and easily removing the background in images or videos

Filed under: Tech Related — Suramya @ 6:29 PM

At times you want to get a copy of the foreground in a given image so that you can use it in another image, however traditionally that required an expertise in photoshop and a lot of patience. Now, thanks to Johnathan Nader there is an opensource tool called backgroundremover that allows you to remove the background from any image quickly and easily.

Installing the tool was quite simple and on my Debian machine I just had to run the following command:

pip3 install backgroundremover --break-system-packages

The –break-system-packages parameter was included because otherwise Debian doesn’t let you install/update global packages using PIP. Another option to install would have been to install a virtual environment for python and update there but I don’t like doing that, and was fairly confident that this wouldn’t break anything important.

After installation the usage is quite simple, you just pass it the image or video where you want to remove the background and the output filename and the system does the rest. It does take a minute or so the first time to download the models etc but after that it is fairly quick to finish. The command to convert looks something like:

backgroundremover -i inputfile.jpg -o outputfile.jpg

Once the script finishes you have an output file without the background. In my testing if the foreground was large and the focus of the image the system was able to successfully remove the background. In some cases a bunch of background artifacts were seen in the image next to the foreground image but it was still quite good.

It would have been awesome if we could have passed a flag to the program to remove the foreground instead of the background. That would have allowed us to remove photo-bombers or random folks who walked in on a photo being taken. I know there are scripts out there that do this but haven’t really tried any of them out.

– Suramya

April 19, 2023

Finally a useful AI Implementation: Making spoken dialog easier to hear in movies and shows

Filed under: Emerging Tech,News/Articles,Tech Related — Suramya @ 6:37 PM

Finally, an AI usecase that is actually useful. There are a ton of use cases where AI seems to be shoehorned in for no reason, but this recent announcement from Amazon about Dialogue Boost which is a new function from that lets you increase the volume of dialogue relative to background music and effects to a consistent volume so you can actually hear the dialog without nearly shattering the eardrums when a sudden explosion happens.
It is something that is still in the testing phase and is only released on some of their products so far. But I am looking forward to it being in general availability.

Dialogue Boost works by analyzing the original audio in a movie or series and identifying points where dialogue may be hard to hear above background music and effects, at which point speech patterns are isolated and audio is enhanced to make the dialogue clearer. The AI targets spoken dialogue rather than a typical speaker or home theater set up that only amplifies the center channel of audio. It’s something that exists on high-end theater set-ups and certain smart TVs, but Amazon is the first streamer to roll out such a feature.

I have gotten used to having subtitles on when I watch something because that ensures that I don’t miss out on any dialogs due to the background music/sounds in the show/movie. This looks like it will alleviate that requirement. I think I will still end up keeping the subtitles on but this will certainly help.

Source: Amazon’s New Tool Adjusts Sound So You Can Actually Understand Movie and TV Dialogue
Announcement: Prime Video launches a new accessibility feature that makes it easier to hear dialogue in your favorite movies and series

– Suramya

April 14, 2023

My app that autoposts to Twitter has been suspended from accessing the Twitter API

Filed under: My Thoughts,Tech Related,Website Updates — Suramya @ 5:44 PM

Yesterday I got an email from Twitter stating the following:

Hello,

This is a notice that your app – Suramya’s Blog – has been suspended from accessing the Twitter API.

Please visit developer.twitter.com to sign up to our new Free, Basic or Enterprise access tiers.

More information can be found on our developer community forums.
Regards,
Twitter Developer Platform

The email actually looks like a really bad phishing email as it has no formatting, doesn’t give any links etc and is just a plain auto-generated email. I almost deleted it as spam but then realized that it could be a notification sent because they are forcing folks to use the new plans. Today I logged in to the Developer account and I was expecting to have an option to select one of the tiers, click save (pay if I was insane and decided to pay) and would be done with it. But that is not the case. I was greeted with the following banner when I logged in:


This App has violated Twitter Rules and policies. As a result, it can no longer be accessed. For assistance, submit a support ticket.

It looks like they couldn’t figure out how to temp block users who need to select a tier before being allowed to continue so they decided to suspend the app instead using the same process as what they would do if the app was suspended for ‘violations of Twitter Rules and policies’. Which is quite amusing because the app been used 12 times in the last 2 months to autopost links to my posts here when I create them. I did use the same app for testing a Twitter export script that I wrote a few months ago but haven’t run it in a while, either.

There is no way for me to edit/choose a tier for my app and I have no interest is spending the time to create another app just to post something on Twitter which will get about 2-10 view on an average. (Usually on the lower end of the scale). This was pretty much the last remaining vestige of my posting on Twitter and I am fine with it not working anymore.. I rather spend that time doing something more productive like watching paint dry.

– Suramya

April 4, 2023

Mastodon is so much better than Twitter, except for its search capabilities

Filed under: My Thoughts,Tech Related — Suramya @ 5:14 PM

Twitter has been slowing becoming less and less useful for getting updates from people you follow. Even my ‘Following’ tab is now showing entries from people I don’t follow and not all posts from the folks I follow show up on their either. Don’t even get me started about the ‘For You’ section which is full of nonsense that I am not really interested in. I have mostly switched over to Mastodon for updates and I see way better engagement over there. My blog auto-posts to both Mastodon and Twitter (along with LinkedIn and Facebook), on Twitter I have 84 followers and 11 followers on Mastodon (I only started posting there in 2023). My Tweets usually get between 2-10 views each and maybe 1 tweet out of 50 will get a response or like. The same post on Mastodon gets a lot more engagement, there have been posts which have had 8-10 replies and multiple likes.

However, that being said one thing that Twitter has which is missing from Mastodon is the ability to search. Earlier today I saw an article on how Twitter seems to have blocked users from authenticating to other services using their SSO offerings. I wanted to learn more about it and tried searching for it on Mastodon, and didn’t get any results (I then tried searching using a hashtag but no luck there as well). So I switched to Twitter and did a search there and immediately I got a lot of results that gave more information on the topic. I am sure that this event is being discussed in Mastodon but it is almost impossible to find because of the way the search is designed.

There is an opt-in project that allows people to opt-in to their setup to allow them to index your toots but because of the ‘amazing’ search in Mastodon, I can’t find the link to the project. 🙁 There are people working on this problem but a extremely vocal minority is hellbent against allowing people to search on Mastodon because they don’t want it. To be fair there are a lot of technical challenges in indexing all the toots across all the instances but it is not an insurmountable problem. It just needs people to look into the problem and others to let them work on the solution.

– Suramya

April 3, 2023

ISRO successfully completes the landing experiment of the Reusable Launch Vehicle-Technology Demonstration

Filed under: Astronomy / Space — Suramya @ 2:07 PM

ISRO (Indian Space Research Organisation) has been quite active in the past few years and has been making significant progress in its mandate to power the next generation of Indian Space capabilities. On Sunday ISRO successfully completed the landing experiment of the Reusable Launch Vehicle-Technology Demonstration (RLV-TD) which is a key part of ISRO’s goal to create a reusable low cost two-stage orbital launch vehicle.

“The RLV took off at 7:10 a.m. by a Chinook helicopter of the IAF as an underslung load and flew at a height of 4.5 km. Once the predetermined pillbox parameters were attained, based on the RLV’s Mission Management and Computer command, the RLV has released mid-air, at a down range of 4.6 km,” ISRO said.

The autonomous landing was carried out under the exact conditions of a Space Re-entry vehicle’s landing —high speed, unmanned, precise landing from the same return path— as if the vehicle arrives from space. Landing parameters such as Ground relative velocity, the sink rate of Landing Gears, and precise body rates, as might be experienced by an orbital re-entry space vehicle in its return path, were achieved. The RLV LEX demanded several state-of-the-art technologies including accurate Navigation hardware and software, Pseudolite system, Ka-band Radar Altimeter, NavIC receiver, indigenous Landing Gear, Aerofoil honey-comb fins and brake parachute system.

In a first in the world, a winged body has been carried to an altitude of 4.5 km by a helicopter and released for carrying out an autonomous landing on a runway. RLV is essentially a space plane with a low lift to drag ratio requiring an approach at high glide angles that necessitated a landing at high velocities of 350 kmph. LEX utilized several indigenous systems. Localized Navigation systems based on pseudolite systems, instrumentation, and sensor systems, etc. were developed by ISRO. Digital Elevation Model (DEM) of the landing site with a Ka-band Radar Altimeter provided accurate altitude information. Extensive wind tunnel tests and CFD simulations enabled aerodynamic characterization of RLV prior to the flight. Adaptation of contemporary technologies developed for RLV LEX turns other operational launch vehicles of ISRO more cost-effective.

Keep in mind that this was a Technology Demonstration test, or what we call Proof of Concept (PoC) in Software Development. Now that we have successfully shown that the technology works as expected, they can start scaling it up, add more functionality/redundancy etc to productionalize it and make it ready for the live test a couple of years down the line.

Looking forward to more such positive news from ISRO going forward.

Source: ISRO successfully conducts the Reusable Launch Vehicle Autonomous Landing Mission

– Suramya

March 18, 2023

Scientists create a working supersolid in the lab

Filed under: Emerging Tech,My Thoughts — Suramya @ 11:34 PM

It seems that every year we learn more about the universe that makes the basic physics that we learned in school inaccurate or rather puts a lot of caveats in to the theories. Originally we had 3 states of matter: Solid, liquid and gas. Then came things like superfluids, Bose–Einstein condensates, quantum spin liquid, supercritical fluid, quark–gluon plasma, Rydberg polaron, and so many more weird possibilities. Last week, scientists from Innsbruck University in Austria have managed to create a new state of matter in 2D called Supersolids. Till now the researchers had only been able to create a 1D (a few molecules long) chain of SuperSolids but using cutting edge research they were able to create a 2D ‘paper’ of supersolid.

If you are like me, by now you will be wondering what on earth is a supersolid… Basically it is a state of matter that incorporates two different states of matter at the same time i.e. it is a solid as well as a superfluid at the same time. This gives it the ability to be a solid and still flow like a liquid without any friction at the same time. If that sounds confusing it is so because we are talking about Quantum effects which seem to exist in a state of constant contradiction and confusion (At least for me, when I try to understand them).

“To picture a supersolid, consider an ice cube immersed in liquid water, with frictionless flow of the water through the cube,” wrote Bruno Labruthe-Tolra, a physicist at Sorbonne Paris North University.

So, to create a supersolid, you first trap some atoms, then cool them, then play with their interactions. “If you tune those correctly, and you tune the shape of the trap correctly, you can get a supersolid,” says Norcia, the lead author.

Using this method, in 2019, researchers began to create a basic, one-dimensional supersolid: essentially, a thin supersolid tube in a straight line.

That’s what Norcia and his colleagues at Innsbruck University and the Austrian Academy of Sciences have now done. By tinkering with the device they used to trap atoms and the process they used to condense the atoms, they were able to extend their supersolid from one dimension into two: from a tiny tube into a small sheet.

There are a lot of interesting usecases for this technology when it matures, we could use it for lubrication in industrial machinery, create frictionless surfaces for tests. It could even be used in vacuum as is for various usecases. But that is still quite a way off because the work to go from 2D to 3D has just started and is still in the pre-research stage. However, while that is going on we do have a superSolid paper available for study while will give us more insight into this fascinating new substance.

The research has been published in Nature: Supersolids go two-dimensional

Source: Popsci.com: We finally have a working supersolid. Here’s why that matters.

– Suramya

March 12, 2023

Researchers create mini-robot that can navigate inside blood vessels and perform surgery autonomously

Filed under: Emerging Tech,Tech Related — Suramya @ 11:13 PM

Performing surgery is a delicate task and at times it is almost impossible to reach the area we want to operate at without having to cut through other important tissues. This is even more apparent when we talk about surgery inside a blood vessel or artery, which could be the key to removing an obstruction or stitch a wound etc. Till now we didn’t have the ability to release an autonomous robot inside a blood vessel that could navigate to the correct location, perform the programmed actions (or allow the doctor to manually take over) and return.

This was only possible in the realm of Science Fiction but thanks to the efforts of Researchers at South Korea’s Hanyang University this is now actually possible in the real world. They have successfully demonstrated that their I-RAMAN (robotically assisted magnetic navigation system for endovascular intervention) robot can travel autonomously to a superficial femoral artery in a pig, deliver contrast dye, and return safely to the extraction point. Their results and paper was published on 9th Feb in IEEE Robotics and Automation Letters: Separable and Recombinable Magnetic Robot for Robotic Endovascular Intervention.

This study presents a separable and recombinable magnetic robot (SRMR) to deliver and retrieve an untethered magnetic robot (UMR) to a target vascular lesion safely and effectively for robotic endovascular intervention. The SRMR comprises a delivery catheter and UMR connected to the end of the delivery catheter by a connecting section. An external magnetic field (EMF) interacts with the permanent magnet of the UMR; it can effectively generate magnetic torque and steer the delivery catheter to reach a target lesion. Furthermore, the rotating EMF allows the UMR of the SRMR to separate from the delivery catheter and perform the tunneling task. After completing the tunneling task, the UMR can be safely recombined with the delivery catheter in the vasculature via a simultaneous application of the EMF and suction force to the delivery catheter. The SRMR functions of steering, separation, movement, tunneling, drug delivery, and recombination are validated in a mimetic vascular model with a pseudo blood clot. Finally, the SRMR is successfully validated in an in vivo experiment of a mini pig’s superficial femoral artery for contrast delivery, separation, movement, and recombination.

This is a fantastic achievement, and although there is a lot of work still left to be done before this can be deployed for actual human use we are still a step closer to truly universal repair bots. Imagine an accident victim who is bleeding internally, the doctor deploys these robots to restitch the blood vessels to stop the internal bleeding and within minutes the bleeding is stopped and the doctor can start the post-op work. I can imagine these being sold as part of the standard medkits in the future (way in the future) where you have a few pre-programmed options available and depending on the situation a person can select the correct option to deploy.

However, all is not rosy (as always). If these go into active use and become common enough to be deployed in med-kits then we would need systems to prevent these bots from being repurposed. For example, instead of being programmed to stitch blood vessels the bots are programmed to cause more damage and start internal bleeding. There are so many other scenarios where this could be misused so we would need to think of all the cases, mitigate the risk and only then deploy them into the world.

That being said, I am still excited to see the possibilities this opens up.

Source: ACM Tech News Newsletter.

– Suramya

« Newer PostsOlder Posts »

Powered by WordPress