iiinova — where your digital self keeps evolving

Adrian Chan @gravity7.bsky.social · 31/03/2026

A recursively self-improving AI: its input is its own output. Its training data is its own generation. Its evaluation is its own judgment. Nothing from outside enters. This isn't exploration. This is confirmation. The most sophisticated echo chamber ever built.

020

Adrian Chan @gravity7.bsky.social · 08/06/2025

Everybody talking about the "new" apple paper might find this MLST interview with @rao2z.bsky.social interesting. "Reasoning" and "inner thoughts" of LLMs were exposed as self-mumblings and fumblings long ago. #LLMs #AI www.youtube.com/watch?v=y1Wn...

youtube.com

Do you think that ChatGPT can reason?

YouTube video by Machine Learning Street Talk

050

Adrian Chan @gravity7.bsky.social · 03/06/2025

yes - people will still need a phone, and a lot of AI products, services, and UI will need a screen. and a touchable one at that.

010

Adrian Chan @gravity7.bsky.social · 21/05/2025

This is interesting, published yesterday. CoT type reasoning shifts attention away from instruction tokens. Paper proposes "constraint attention" to keep models attentive to instructions when doing CoT. #AI #LLM www.arxiv.org/abs/2505.11423

arxiv.org

When Thinking Fails: The Pitfalls of Reasoning for Instruction-Following in LLMs

Reasoning-enhanced large language models (RLLMs), whether explicitly trained for reasoning or prompted via chain-of-thought (CoT), have achieved state-of-the-art performance on many complex reasoning ...

030

Adrian Chan @gravity7.bsky.social · 14/05/2025

Clarifying questions w #LLMs increase user satisfaction when users can see the point of answering them. Specific questions beat generic ones. But I wonder if this changes when #agents are personal assistants, & are more personal & more aware. #UX #AI #Design arxiv.org/abs/2402.01934

arxiv.org

Clarifying the Path to User Satisfaction: An Investigation into Clarification Usefulness

Clarifying questions are an integral component of modern information retrieval systems, directly impacting user satisfaction and overall system performance. Poorly formulated questions can lead to use...

000

Adrian Chan @gravity7.bsky.social · 14/05/2025

Interesting - could #LLMs in search capture context missed when googling? "backtracing ... retrieve the cause of the query from a corpus. ... targets the information need of content creators who wish to improve their content in light of questions from information seekers." arxiv.org/abs/2403.03956

arxiv.org

Backtracing: Retrieving the Cause of the Query

Many online content portals allow users to ask questions to supplement their understanding (e.g., of lectures). While information retrieval (IR) systems may provide answers for such user queries, they...

000

Adrian Chan @gravity7.bsky.social · 14/05/2025

They mostly test whether they can steer pos/neg responses. But given Shakespeare was a test also, wld be interesting to extract style vectors from any number of authors then compare generations. (Is this approach used in those "historical avatars?" No idea.)

000

Adrian Chan @gravity7.bsky.social · 14/05/2025

@tedunderwood.me In case you haven't seen this paper, you might find interesting. Researchers extract style vectors (incl from Shakespeare) and apply to an LLM internal layers instead of training on original texts. Generations can then be "steered" to a desired style. arxiv.org/abs/2402.01618

arxiv.org

Style Vectors for Steering Generative Large Language Model

This research explores strategies for steering the output of large language models (LLMs) towards specific styles, such as sentiment, emotion, or writing style, by adding style vectors to the activati...

110

Adrian Chan @gravity7.bsky.social · 14/05/2025

But design will need to focus on tweaking model interactions so that they track conversational content and turns over time. For example with bi-directional prompting: models prompt users to keep conversations on track. This seems a rich opportunity for interaction design #UX #IxD #LLMs #AI

010

Adrian Chan @gravity7.bsky.social · 14/05/2025

to sustain dialog. Social interaction face to face or online is already vulnerable to misunderstandings and failures, and we have use of countless signals, gestures, etc w which to rescue our interactions. A communication-first approach to LLMs for conversation makes sense, as talk is not writing.

100

Adrian Chan @gravity7.bsky.social · 14/05/2025

"when LLMs take a wrong turn in a conversation, they get lost and do not recover." Interaction design is going to be necessary to scaffold LLMs for talk, be it voice or single user chat or multi-user (e.g. social media). It's one thing to read/summarize written documents, quite another ...

100

Adrian Chan @gravity7.bsky.social · 14/05/2025

"LLMs tend to (1) generate overly verbose responses, leading them to (2) propose final solutions prematurely in conversation, (3) make incorrect assumptions about underspecified details, and (4) rely too heavily on previous (incorrect) answer attempts." arxiv.org/abs/2505.06120

arxiv.org

LLMs Get Lost In Multi-Turn Conversation

Large Language Models (LLMs) are conversational interfaces. As such, LLMs have the potential to assist their users not only when they can fully specify the task at hand, but also to help them define, ...

100

Adrian Chan @gravity7.bsky.social · 14/05/2025

"LLMs ... recognize graph-structured data... However... we found that even when the topological connection information was randomly shuffled, it had almost no effect on the LLMs’ performance... LLMs did not effectively utilize the correct connectivity information." www.arxiv.org/abs/2505.02130

arxiv.org

Attention Mechanisms Perspective: Exploring LLM Processing of Graph-Structured Data

Attention mechanisms are critical to the success of large language models (LLMs), driving significant advancements in multiple fields. However, for graph-structured data, which requires emphasis on to...

000

Adrian Chan @gravity7.bsky.social · 12/05/2025

Perhaps one could fine tune on Lewis Carroll, then feed the model with philosophical paradoxes, and see whether the model produces more imaginative generations.

000

Adrian Chan @gravity7.bsky.social · 12/05/2025

I think because this isn't making the model trip, synesthetically, but is simply giving it juxtapositions. So what is studied is a response to these paradoxical and conceptually incompatible prompts, not a measure of any latent conceptual activations or features.

100

Adrian Chan @gravity7.bsky.social · 12/05/2025

Let's dose an LLM and study its hallucinations! LLMs were fed "blended" prompts, impossible conceptual combinations, meant to elicit hallucinations. Models did not trip, but instead tried to reason their way through their responses. arxiv.org/abs/2505.00557

arxiv.org

Triggering Hallucinations in LLMs: A Quantitative Study of Prompt-Induced Hallucination in Large Language Models

Hallucinations in large language models (LLMs) present a growing challenge across real-world applications, from healthcare to law, where factual reliability is essential. Despite advances in alignment...

110

Adrian Chan @gravity7.bsky.social · 10/05/2025

Yes and the label applied says as much about the person as it does about the model. In the world of creatives, the most-used term now is "slop," derived perhaps from enshitification. The latter capturing corporate malice where the "slop" is AI-generated byproduct unfit for human consumption...

010

Adrian Chan @gravity7.bsky.social · 10/05/2025

Thread started w your second post so yes I missed the initial post. Never mind.

000

Adrian Chan @gravity7.bsky.social · 10/05/2025

Assuming alignment using synthetic data is undesirable, one route is to complement global alignment (alignment to some "universally" preferred human values) w local, contextualized alignment, via feedback and use by the user. Tune the LLM's behavior to user preferences.

100

Adrian Chan @gravity7.bsky.social · 10/05/2025

Customized LLMs use the feedback obtained from the individual user interactions and align to those.

100

Adrian Chan @gravity7.bsky.social · 10/05/2025

Staying power of ceasefires becoming a proxy for multilateral resilience amid baseline rivalries?

010

Adrian Chan @gravity7.bsky.social · 10/05/2025

I think this will be one accelerant for individualized/personally customized AI - e.g. personal assistants. The verifiers can use the user's preferences and tune to those rather than apply globally aligned behavioral rules.

100

Adrian Chan @gravity7.bsky.social · 10/05/2025

It's also a problem of use cases and user adoption. Though it may turn out that Transformer-based AI does indeed fail to meet expectations. There's a lot of misunderstanding and anthropomorphism of AI's reasoning, for example, that might not turn out well.

010

Adrian Chan @gravity7.bsky.social · 10/05/2025

Coincidentally many startups of that time set up in loft & warehouse spaces w exposed concrete & steel beams.... I like this analogy especially for Social Interaction Design/Social UX, where "social architecture" is exposed for users to take up in norms, behaviors, expectations for how to engage

020

Adrian Chan @gravity7.bsky.social · 06/05/2025

I can't disagree w that. Reflection through reading employs more critical thinking skills than conversation; bots solicit unserious interaction & even attempts to "hack" guardrails. I'm a huge reader but I do have lengthy convos w ChatGPT, likely because I read/reflect.

100

Adrian Chan @gravity7.bsky.social · 06/05/2025

Agree w you. Tariffs as targeted protections of domestic ind, as reciprocity, as reshoring incentives, as embargoes - all these are different & neglect unintended consequences as we're seeing in markets & bonds & dollar. Regardless of motives it's now a matter of game theory - who moves, when, etc

020

Adrian Chan @gravity7.bsky.social · 06/05/2025

For now I can see that chatbots likely would fail to provide accurate or probable reasoning if prompted for explanations of historical choices, actions, etc, for lack of proper historical context. But this too could be improved w training on secondary lit. It's admittedly all rather Black Mirror.

000

Adrian Chan @gravity7.bsky.social · 06/05/2025

To learn a historical figure from a book however is to imagine their reasons, motives, actions in abstract. (Which is fine.) To have them personified as chatbots seems absurd and kitschy - but might reach some students who simply don't engage by reading.

200

Adrian Chan @gravity7.bsky.social · 06/05/2025

These bots likely are built on texts but not graphs, w which they could better hew to facts, etc. They might be trained to better interact - but this would be to layer pedagogical learning methods onto the bot's conversation style (is still interesting). On moral view, you're absolutely right.

120

Adrian Chan @gravity7.bsky.social · 06/05/2025

So bureaucratic politics would explain the apparent incoherence as members of cabinet vie for position, favor, leadership and access. A rational approach might have started cuts only after policies had been set.

100

Adrian Chan @gravity7.bsky.social · 06/05/2025

I think unintended consequences of poorly reasoned, hastily implemented cuts (DOGE), lack of policy coordination (Navarro, Lutnick, Bessent), for pol (Rubio, Hegseth, Vance), & psychic intents of tariffs. Also in China's case tariffs not about grievance only but also domestic reshoring.

100

Adrian Chan @gravity7.bsky.social · 06/05/2025

Perhaps v2, v3, etc will be better? I haven't used them, not here to defend them. But I could see benefits of students becoming more interested because mode of interaction is more compelling than books. If historical bots were tuned to prompt students for different questions, etc?

100

Adrian Chan @gravity7.bsky.social · 05/05/2025

And language fools people, period. I think the issue of style, which AI can do so incredibly well, is fascinating. We're soon close to being unable to tell authentic from AI generated apart. Burden then shifts to culture & its tastes & discernment to decide what value/or not to put on AI content

010

Adrian Chan @gravity7.bsky.social · 05/05/2025

We tolerate bad & even fawn over the tasteless. Agree. AI output will become indistinguishable from the real thing starting w text then moving up the sensory/medium ladder. Is why document summaries etc r so "easy." At same time our culture does risk pollution by kitsch, esp on social platforms.

010

Adrian Chan @gravity7.bsky.social · 05/05/2025

Agree, but do also think that our tastes, culturally speaking, leave us vulnerable to AI generated content/fakes etc that we won't catch. I think that's coming soon as quality of AI output becomes indistinguishable. Parallel wld be counterfeits in art world: r they not art even if they're copies?

110

Adrian Chan @gravity7.bsky.social · 05/05/2025

except the Black Mirror ones ;-)

020

Adrian Chan @gravity7.bsky.social · 05/05/2025

1 - true now; 2 - true always; 3 - AI's shortcomings wrt cultural product counterfeits/imitations/mimicry/novelties aren't on acct of no world model, but no social model, & more importantly, that social validation of human cultural prods still celebrates authenticity. AI fakery is socially rejected.

010

Adrian Chan @gravity7.bsky.social · 04/05/2025

This chapter in RLHF demonstrates that there's a zone of indeterminacy between generalized human feedback and individualized human feedback, in which the social presence of AI in particular is especially vulnerable.

000

Adrian Chan @gravity7.bsky.social · 04/05/2025

The more personal the model, the more likely it mines prior convos for style, word choice, and like a good therapist should use some mirroring to increase familiarity. AI design will need heuristics for model "personhood" and proximity. (the movie "Her)

100

Adrian Chan @gravity7.bsky.social · 04/05/2025

In human social interaction, these cues build trust, interest, sustain attention, etc. They conflate personal affirmation and substantive agreement - hence they are doubly effective but also doubly ambiguous. "Good idea!" is affirming & possibly agreement/approval.

100

Adrian Chan @gravity7.bsky.social · 04/05/2025

An unsurprisingly sharp deep dive into sycophancy & RLHF. Whilst this episode has its tech explanations, the social interaction aspects are unsolved & will rise as models are personalized, as mirroring is an effective design feature, & we're not good at distinguishing affirmation from agreement.

110

Adrian Chan @gravity7.bsky.social · 04/05/2025

Analogous to honing in on the solution to a problem, this #LLM research draws conclusions from the most common "reasoning trace" & compares to final answer. Overthinking by reasoning models can be a waste of tokens - but which tokens? #AI www.arxiv.org/abs/2504.20708

arxiv.org

Beyond the Last Answer: Your Reasoning Trace Uncovers More than You Think

Large Language Models (LLMs) leverage step-by-step reasoning to solve complex problems. Standard evaluation practice involves generating a complete reasoning trace and assessing the correctness of the...

010

Adrian Chan @gravity7.bsky.social · 03/05/2025

"Opportunistic sycophancy?" Anthropic's discovery of an influence bot network shows how conversation can be weaponized. This isn't the Cambridge Analytica-styled user preference targeting, but language/conversation instead. #AIethics www.anthropic.com/news/detecti...

anthropic.com

Detecting and Countering Malicious Uses of Claude

122

Adrian Chan @gravity7.bsky.social · 03/05/2025

Wld b interesting to apply this resrch to multi-turn convos w AI - AI can't anticipate convo as we do, as it can't anticipate "where we are going" - thus the temporal anticipation on AI's side of convo will lack look ahead phrases.

000

Adrian Chan @gravity7.bsky.social · 03/05/2025

"coherent discourse organisation. This is achieved by either pointing backward to previously discussed material or forward to upcoming propositions" ...wouldn't lack of cataphoric be explained by Transformer architecture, so not anticipating arguments made later?

000

Adrian Chan @gravity7.bsky.social · 02/05/2025

This prompted the idea of using mechanistic probing on a model trained on period literature to conduct a digital Foucauldian examination of latent sociocultural features (as evidenced by latent layer features, activitations, circuits etc). ;-) Period models - Black Mirror Hotel Reverie episode?

010

Adrian Chan @gravity7.bsky.social · 02/05/2025

One potential but not primary benefit to personalizing LLMs to user preferences (memory & sycophancy) might be to create more signal for model verifiers & reasoning - but is there a risk then we get individualized AI-generated filter bubbles? Cognitive biases mirrored back to us?

040

Adrian Chan @gravity7.bsky.social · 02/05/2025

"framed as answers" includes many types of response, as there are many types of questions. An "answering" response may be confirmation, fact, recommendation, suggestion, advice, contradiction/negation/challenge, etc. Search is frustrating bec it can't adequately interpret intent behind the query.

000