Anant's Social Feed

Funny scenario which may or may not represent current it world.

Imagination:

Lets use modules in code, so that one single place can be fortified with all precautions.

Third party developer finds a flaw fixes it everyone else uses the updated version world is instantly more secure.

Reality:

Lets use modules as it saves time for development less headache for us.

We dont fully trust the third party to keep going in same direction as us so freeze the version.

News about oss devs modifying code against policies is popularized. Third party dependency tool companies take it to next level spread the news you cant trust modules.

New approaches from version freezing to hashbased mapping comes into picture reducing risk as version could be easily over written but not code commit hash.

A bug appears in code, author fixes it, everyone using the module or atleast those who are aware fix it, anyone using those softwares update their code bases, this keeps cascading down to lower levels.

And may be just may be in next 5 years the world is instantaneous better secured. /s

Who is at fault, what can be done to reduce unnecessary loops, is fixating on third party dependencies the right approach, should the focus be somewhere else? Some interesting questions with no clear answers

Using OpenAI chatgpt api trained on your own notes seems like an interesting idea.

Repo : https://github.com/brianpetro/obsidian-smart-connections

Equally polarized reactions on reddit: https://www.reddit.com/r/ObsidianMD/comments/11s0oxb/chat_with_your_notes_now_available_in_the_smart/

I have my concerns around sharing data with openai althought openai has made it pretty clear that they are "no longer" using data provided by API for training the overall model (since 1st march 2023). https://openai.com/policies/api-data-usage-policies

I have some pretty interesting things in my repository that; if i could get a second opinion on i would have loved to share with humans. But humans as they go tend to form judgements on other people. thats where AI can help but the current state of Generative AI is far from being really able to help rather then spouting some nonsense which may make sense may derail my whole thought process.

Hopefully one day we will have opensource local only kind of model where if i dont like what it says i would be in a position to format it all by myself.

#PKM #personalnotes #privacy #localfirst

SCRCPY : screen mirroring app for android : v2 released with audio support:
https://blog.rom1v.com/2023/03/scrcpy-2-0-with-audio/

The Center Of Our Galaxy Smells Like Raspberries And Tastes Like Rum | IFLScience

https://www.iflscience.com/center-our-galaxy-smells-raspberries-and-tastes-rum-24770

#space

@ondrej Hey it seems you broke your redirect from sury.org. Cant find you via your officially listed name found you via @ondrej

Finally found a save #Mastodon post in #Obsidian plugin

https://github.com/DominikPieper/obsidian-ReadItLater

This does the job a bit kludgy but it does the job for now atleast for mastodon not sure of whole #fediverse

#PKM #dataHoarders

cc: @eleanorkonik

Here is a thought about #ChatGPT or likes (#GenerativeAI) usage in #Academics.

#Teachers should encourage #students to ask questions to it and find ways to get the right answer. Teachers ask them to learn about a topic using chatGPT and then instead of teaching students about that topic using old methods clear the doubts in live class. let each student share what they learned for 30 minutes and then in next 30 minutes confirm what is right and confirm what is wrong and if marked as wrong then why it is wrong explain and if required prove.

This way

1. You are using #ChatGPT as a #tool not an enemy just like we treat calculator now.

2. AI Systems have limitations they dont have all answers and right questions can make them fumble. Teachers need to ask the right question to judge the learning not the Rota of students.

3. AI Systems allow you to ask questions in isolation without looking stupid.Answer could be right or wrong, and thats where teachers come and they should help students, discarding it as a FAD or embarcing it fully with open arms is both wrong.

In the end all i will say is (In classic academic rota essay format) AI is good tool but a bad master. Focus on leveraging it rather then it leveraging you.

What do you think?

A major change that I see between using google for searches vs using chatgpt to find answers.

In search engine you want to be as creative by reducing input to bare minimum so as to not pollute the search results and specific keywords are given. Where as in chatGPT you go as much in detail as you can to narrow the result down.

Why I asked myself:

1. ChatGPT is assumed to be a language understanding system so we converse with it. Search engine are software so we give instructions pin pointed focused.
2. Search engines have been gamed for many years in the name of SEO or what not. AI's are yet to be massively gamed and hence more details allow you to narrow the results.
3. Search engines give you options. AI gives you answer. you want that one shot to be as accurate as you can get.

Ya ya, i am right now fully ignoring the aspect of assuming AI as human or that AI are bad etc. I am simply making an observation that i have seen recently.

The end of writing https://ia.net/topics/the-end-of-writing-ia-on-ai by @iawriter@mastodon.online

It comes as a surprise to me a lot of times that there is a big population in U.S.A. that assumes anything happening in USA is a world interest thing and without specifying context they can talk about it and everyone will get the context.

However if god forbid same assumption is made by an asian or an african they are pin pointed that this is local to you please state context.

Feels like screaming at times, You know there is a whole big earth with many more continents then just that one country.

#RespectOthers #nations