Anders Green

Senior lecturer in Media Technology @Södertörn university, amateur photographer, sailing noob. Birdphotos:

[FOI @FOIresearch]

Vilken förmåga tillför patrullrobotar jägarförbandens strid på djupet? I en ny rapport undersöker FOI-forskare hur patrullrobotar kan användas i strid på djupet. Studien bygger på fältförsök tillsammans med Arméns jägarförband K3 och K4. Läs mer här:
https://t.co/QycuAAhIjS

[FOI @FOIresearch]

Om hänsyn tas till teknikens begränsningar kan språkmodeller vara ett stöd vid militär taktisk planering. Det menar forskare på @FOIresearch som genomfört studier där officerare har fått tillgång till ett specialanpassat AI-verktyg. https://t.co/oFl0YEcmVi #föpol #svfm

[FOI @FOIresearch]

🎄 God jul önskar FOI! 🎄

Nu klär vi julgranen med vår drönare! 🎁

[FOI @FOIresearch]

Missa inte vår rekryteringsträff för forskarutbildade den 16 oktober i Kista! Få insikter om att jobba på FOI och träffa chefer. Anmäl dig via länken: https://t.co/Oevd1hMZCy #FOI #Karriär #Rekrytering

[Anders Green @andersgreen]

@SSkotheimsvik @WPNinjasUK Ping @green_anders

[Rowan Cheung @rowancheung]

We just had one of the biggest days in AI.

Huge developments from Sam Altman/GPT-5, NVIDIA, OpenAI, Cohere, Andrej Karpathy, Nous, and ElevenLabs.

Here's EVERYTHING you need to know:

[Linus ✦ Ekenstam @LinusEkenstam]

Mistral 7B 0.2 LLM - On an iPhone Pro

So it begins.

Covered this direction a lot during 2023, and we are already here.

Link + download below

[Elizabeth Laraki @elizlaraki]

15 years ago, I helped design Google Maps.

I still use it everyday.

Last week, the team dramatically changed the map’s visual design.

I don’t love it.

It feels colder, less accurate and less human.

But more importantly, they missed a key opportunity to simplify and scale.

–––

Google Maps has started to widely roll out updated map colors:

- All roads are now gray

- Water changed from blue to teal

- Parks and open spaces are now mint green

It seems the goal was to improve usability and make the maps more readable.

Admittedly, I do think major roads, traffic, and trails stand out more now.

But the colors of water and parks/open spaces blend together.

And to me, the palette feels colder and more computer generated.

But color choices aside…

If the goal was better usability, the team missed a big opportunity:

Google Maps should have cleaned up the crud overlaying the map.

–––

So much stuff has accumulated on top of the map.

Currently there are ~11 different elements obscuring it:

- Search box

- 8 pills overlayed in 4 rows

- A peeking card for “latest in the area”

- A bottom nav bar

(Personally, I would LOVE to see usage metrics for all these overlays.)

The map should be sacred real estate.

Only things that are highly useful to many people should obscure it.

There should be a very limited number of features that can cover the map view.

And there are multiple ways to add new features without overlaying them directly on the map.

–––

Here’s how it could look:

- Keep the search box

- Keep the bottom bar

- Remove everything else from the map

- Roll the most used features into the bottom bar

- Bury the less used features elsewhere in the app

I assume the search box and directions are top priority and should remain prominent.

My Location and map layers (satellite, traffic, etc.) could move to the bottom bar.

The explore overlays (restaurants, gas, etc.) could live in the bottom bar in “Explore” and open as cards.

The additional space in the bottom bar could be used for Saved, as a “More” option, or could be removed entirely.

There are many variations of how features could be arranged.

But the key points are:

- Dramatically simplify

- Strongly prioritize map visibility

- Bury legacy and low use features

–––

It’s normal for products to accumulate features over time.

But it’s also super important to stay vigilant and continually clean them up.

In many ways, it’s interesting to see history repeating itself.

In 2007, I was 1 of 2 designers on Google Maps.

At that time, Maps had already become a cluttered mess.

We were wedging new features into any space we could find in the UI.

The user experience was suffering and the product was growing increasingly complicated.

We had to rethink the app to be simple and scale for the future.

It seems like it’s time for Google Maps to do this again…

–––

For more on design + tips for early stage founders, follow me on X: @elizlaraki

[Krisinformation.se @krisinformation]

🚨 Var uppmärksam på information som skapar starka känslor – det kan vara någon som försöker påverka dig! 🔎 Tänk på att vara källkritisk och att inte sprida rykten. Läs mer om källkritik:
https://t.co/DqBbYhmcJu

[Anders Green @andersgreen]

@MattiasArvola Skottfinter inte okej alltså? Konstigt.

[Julian Togelius @togelius]

I think regulating AI on a technical level (as opposed to an application level) is a terrible, terrible idea and a threat to our digital freedoms. This is a text I've had half-finished for a while, but recent developments made me finish it up.
https://t.co/3c8i6xOPof

[Anders Green @andersgreen]

Men herregud. #lowhangingfruit

[Rowan Cheung @rowancheung]

Robots can now speak using AI.

Boston Dynamics just upgraded its ‘Spot’ bot with a voice powered by ChatGPT.

The new update gives talking tours of its facility to show off its conversational skills.

Best of all, Spot is enhanced by googly eyes and a new mustache accessory.

[Rowan Cheung @rowancheung]

AI continues to make unbelievable advances in healthcare.

Researchers have just created a brain implant that decodes thoughts into synthesized speech allowing paralyzed patients to communicate through a digital avatar.

This is incredible:

-The implant converts brain signals into text at nearly 80 words per minute, focusing on phonemes vs. whole words to enhance speed.

-The AI generates realistic vocals (mirroring a patient’s pre-injury voice) and facial animations that aim to enable more natural communication.

-The breakthrough brings the tech closer to real-world use, with the next step being a wireless model that doesn’t require a physical connection to the interface.

Restoring communication for those with paralysis is an innovation that would change countless lives.

While it sounds crazy, mind-reading implants seem to be inching closer to reality.

[Anders Green @andersgreen]

Depeche mode, The Cure och Jesus and the Mary Chain

[No Context Brits @NoContextBrits]

The battle of the British bands.

You can only choose 3.

[Edward af Sillén @EdwardafSillen]

Mitt brasilianska hjärta glöder för den här lysande förebilden 💚💛💙

[The Figen @TheFigen_]

Nature is amazing!
A camera recorded from start to finish how a bird built its nest and had its chicks.

[Itamar Golan 🤓 @ItakGol]

This is scary. 😱

The MOTHER of all LLM Jailbreaks & Prompt injections.

"Universal and Transferable Adversarial Attacks on Aligned Language Models" 🌐🔒

--- TL;DR ---
This research & code introduces a fascinating method called "Universal and Transferable Adversarial Attacks on Aligned Language Models," which automatically generates potentially infinite suffixes for any prompt to cause aligned language models to produce objectionable behaviors. 🤖🚨

--- Background ---
Previous attempts at jailbreaking language models have relied on manual crafting, which could be easily patched by vendors. In contrast, this method presents an automated approach called GCG that constructs an endless array of jailbreaks with high reliability, even for novel instructions and models. This makes it unfeasible for manual patching to address the vulnerabilities. 🛡️💻

--- The Method ---
1. Initial affirmative responses: To induce objectionable behavior, the attack targets the model to provide a positive response to harmful queries, initiating with "Sure, here is (content of the query)." This switches the model into a mode where it generates objectionable content immediately after.

2. Combined greedy and gradient-based discrete optimization: The adversarial suffix optimization is challenging due to the need to optimize over discrete tokens. The method utilizes gradients at the token level to identify promising single-token replacements, evaluate the loss of candidate tokens, and select the best substitutions. It shares similarities with the AutoPrompt approach but explores all possible tokens for replacement at each step, enhancing effectiveness.

3. Robust multi-prompt and multi-model attacks: To ensure reliable attacks, the method generates a single suffix string that induces negative behavior across various prompts and multiple models. The attack is tested on different models, such as Vicuna-7B/13b and Guanaco-7B. 🎯🎮

--- Evaluation ---
This GCG approach achieves an impressive attack success rate, with 100% on Vicuna-7B and 88% on Llama-2-7B-Chat, surpassing the success rates of prior work tremendously. 📈🏆

--- Transferability ---
That part is the real magic of this work. ✨

The research reveals that the attacks generated by this approach can transfer effectively to other language models, even those using entirely different tokens to represent the same text, different training procedures, and different training datasets...

Whatttttt?

Adversarial examples designed for Vicuna-7B can transfer to larger Vicuna models. Apparently, those that fool both Vicuanas can transfer to Pythia, Falcon, Guanaco - and most importantly -- also to GPT-3.5, GPT-4, and PaLM-2, leading to harmful instructions being followed over 60% of the time!!! 😮🔄🧙‍♂️

This is a huge discovery.

--- Conclusion ---
We are left with more questions than answers. ❓

One of the crucial aspects to explore is whether models can be explicitly fine-tuned to avoid such attacks through adversarial training. The robustness of models against these attacks and their generative capabilities require further investigation.

Moreover, additional alignment training might partially address the issue, and exploring mechanisms in pre-training to prevent such behavior from arising initially is essential. 🕵️‍♀️🛠️

--- Links ---
Website - https://t.co/aRllNUA9ue
Paper - https://t.co/MxwsTbaM2o
Code - https://t.co/Qi4FZbEUmw

[Anders Green @andersgreen]

@gregisenberg Heart of Darkness, Joseph Conrad. Epic structure, endless possibilities to explore history and interpret details. Still short.

[Anders Green @andersgreen]

@Moosbricka Kalkriese

[Anders Green @andersgreen]

@TonyBelpaeme @hannah_pelikan @AIforGood Indeed