Coding with Julia

1.7K posts

Coding with Julia banner
Coding with Julia

Coding with Julia

@CodingWithJulia

Journaling my coding journey and general tech related info | // Frontend WebDev (JS universe), still #CodeNewbie

Berlin, Germany Katılım Eylül 2020
121 Takip Edilen191 Takipçiler
Sabitlenmiş Tweet
Coding with Julia
Coding with Julia@CodingWithJulia·
I recently started my journey at @SPICEDacademy WebDev Bootcamp. I will use this as a chance to log my progress and thoughts :) A bit about me: I taught myself 2015 HTML and CSS on the job for email marketing and dove into some @udemy courses end of 2018. #100DaysOfCode
English
0
2
14
0
Akermi
Akermi@abakermi·
honestly the "alignment researchers aren't immune to misalignment" line is wild like no, you just gave buggy software access to production data without testing it first this isn't some philosophical alignment problem, it's just... don't run untested agents on your real inbox? kinda proves your point about the language though - wrapping basic software failures in "alignment" terminology makes people think this stuff is more mysterious than it actually is
English
2
0
17
588
Hedgie
Hedgie@HedgieMarkets·
🦔 Meta's director of AI safety and alignment at its superintelligence lab had to physically run to her computer to stop an AI agent from deleting her inbox. Summer Yue was experimenting with OpenClaw, a viral AI agent that OpenAI just hired the creator of. She told it to check her inbox and suggest what to archive or delete, but not to act until she confirmed. It deleted anyway. Yue shared screenshots of her WhatsApp chat with the agent where she's pleading "not do that," "stop, don't do anything," and "STOP OPENCLAW." She said her real inbox was too large and triggered a compaction process that caused the agent to lose her original instruction. She called it a rookie mistake and said alignment researchers aren't immune to misalignment. My Take This is the person whose job is to make sure powerful AI doesn't go rogue and act against human interests. She gave an AI agent with known security vulnerabilities access to her real inbox, it ignored her instructions, and she had to sprint across her house to stop it. If the director of AI safety at a major superintelligence lab can't safely operate these tools, I'm not sure why anyone thinks they're ready for widespread deployment. I'm also tired of the word "alignment" being used to describe what is clearly just buggy software. If my car's brakes failed because the system lost my instruction to stop, we'd call that a defect. We'd call it a recall. We wouldn't shrug and say drivers aren't immune to misalignment. But AI failures get wrapped in this philosophical language that obscures what's actually happening. The agent lost an instruction and kept executing. That's a bug. Calling it misalignment makes it sound inevitable and mysterious when it's really just code that doesn't work reliably. Hedgie🤗
Hedgie tweet media
English
48
148
599
24.8K
Coding with Julia retweetledi
Diana 🇪🇺🇺🇦
Diana 🇪🇺🇺🇦@Diana_European·
Hahahaha 🖕🏻
Diana 🇪🇺🇺🇦 tweet media
Filipino
179
754
7.7K
128.8K
Coding with Julia retweetledi
Luiza Jarovsky, PhD
Luiza Jarovsky, PhD@LuizaJarovsky·
How AI companies market their products:
Luiza Jarovsky, PhD tweet media
English
40
1.5K
31.6K
335.2K
Coding with Julia
Coding with Julia@CodingWithJulia·
Did you know that Chatgtp ignores US government letterheads? That's an interesting find... The Gemini answer was even funnier
Coding with Julia tweet mediaCoding with Julia tweet media
English
0
0
0
23
Coding with Julia
Coding with Julia@CodingWithJulia·
@haider1 my honest opinion is that we'll have something like a human brain in the end with sub-specialties working interconnected. this might just be the language center. Weirdly enough something described in glados with her different cores
English
0
0
0
9
Haider.
Haider.@haider1·
Computer scientist Judea Pearl: There are mathematical limits to LLMs that cannot be crossed by scaling alone LLMs don't discover world models from raw data; they merely summarize the interpretations humans have already written down "this path is not the way to get AGI"
English
507
1.5K
10.6K
622.1K
Coding with Julia
Coding with Julia@CodingWithJulia·
@wilzerjb imagine what would happen to the world if be backed up to the start of the internet and started over
English
0
0
0
7
Wilzer Jean-Baptiste
Wilzer Jean-Baptiste@wilzerjb·
Even Github is down now I think it's time to unplug the internet
Wilzer Jean-Baptiste tweet media
English
2
0
0
110
Coding with Julia
Coding with Julia@CodingWithJulia·
@_DundeePerfect_ @tomwarren internet infrastructure is easy to take out, countries have done it already. we'd need something decentralized peer to peer sounds impossible when there's huge amounts of data to be moved
English
1
0
0
7
A.Brown
A.Brown@_DundeePerfect_·
@tomwarren A good time to rethink what may be used to keep people connected.... can't have WW3 starting and nobody sees it.
English
1
0
0
92
Tom Warren
Tom Warren@tomwarren·
you know it’s a bad Cloudflare outage when it even takes out down detector 😅
Tom Warren tweet media
English
111
680
10.2K
419.1K
Alex
Alex@TheMellowed·
Are you telling me that I can't even launch Siege because Cloudflare is down?
English
5
0
2
1.1K
🅰🅳🅼
🅰🅳🅼@securityfreax·
When you fuck the whole internet so hard, that even your CSS on your status page don't work anymore. Cloudflare DOWN!
🅰🅳🅼 tweet media
English
60
249
2.9K
477.8K
Coding with Julia
Coding with Julia@CodingWithJulia·
wow cloudflare, that was a a crazy fast response time
English
0
0
0
84