headalgorithm

Page 2 for me. Everything looks fine

It might be down to this:

https://x.com/embirico/status/2021376881942200801

I would love to see this kind of thing for hn comments.

"Certain grammatical rules never appear in any known language. By constructing artificial languages that have these rules, linguists can use neural networks to explore how people learn."

If only we could replace the real Elon

This Michael O Church? https://michaelochurch.wordpress.com/2017/12/05/why-i-didnt-...

Previously on HN: https://news.ycombinator.com/item?id=39630759

Life is short - Don't waste it

Current discussion: https://news.ycombinator.com/item?id=36551941

Some context: https://writings.stephenwolfram.com/2023/06/prompts-for-work...

Some recent and related discussions:

https://news.ycombinator.com/item?id=36119920

https://news.ycombinator.com/item?id=36056488

https://news.ycombinator.com/item?id=35273707

https://news.ycombinator.com/item?id=35264965

https://news.ycombinator.com/item?id=35242458

Recent discussion: https://news.ycombinator.com/item?id=35985240

Link to paper: https://arxiv.org/abs/2305.16075

I love all this stuff, but I have to agree there is so much GPT chatter that it drowns out everything else. But it will pass, like everything before it.

Abstract:

A major risk of using language models in practical applications is their tendency to hallucinate incorrect statements. Hallucinations are often attributed to knowledge gaps in LMs, but we hypothesize that in some cases, when justifying previously generated hallucinations, LMs output false claims that they can separately recognize as incorrect. We construct three question-answering datasets where ChatGPT and GPT-4 often state an incorrect answer and offer an explanation with at least one incorrect claim. Crucially, we find that ChatGPT and GPT-4 can identify 67% and 87% of their own mistakes, respectively. We refer to this phenomenon as hallucination snowballing: an LM over-commits to early mistakes, leading to more mistakes that it otherwise would not make.

@Christopher: I think your domain chrbutler.com is shadowbanned. I vouched for this post, but you have a lot of posts that are marked dead.

Recent discussion: https://news.ycombinator.com/item?id=36006561

Abstract:

Artificial agents have traditionally been trained to maximize reward, which may incentivize power-seeking and deception, analogous to how next-token prediction in language models (LMs) may incentivize toxicity. So do agents naturally learn to be Machiavellian? And how do we measure these behaviors in general-purpose models such as GPT-4? Towards answering these questions, we introduce MACHIAVELLI, a benchmark of 134 Choose-Your-Own-Adventure games containing over half a million rich, diverse scenarios that center on social decision-making. Scenario labeling is automated with LMs, which are more performant than human annotators. We mathematize dozens of harmful behaviors and use our annotations to evaluate agents' tendencies to be power-seeking, cause disutility, and commit ethical violations. We observe some tension between maximizing reward and behaving ethically. To improve this trade-off, we investigate LM-based methods to steer agents' towards less harmful behaviors. Our results show that agents can both act competently and morally, so concrete progress can currently be made in machine ethics--designing agents that are Pareto improvements in both safety and capabilities.

https://archive.is/8Kx4v

Previous discussion: https://news.ycombinator.com/item?id=15897809

"The AI Incident Database is dedicated to indexing the collective history of harms or near harms realized in the real world by the deployment of artificial intelligence systems. Like similar databases in aviation and computer security, the AI Incident Database aims to learn from experience so we can prevent or mitigate bad outcomes."

Abstract:

Benchmark datasets play a central role in the organization of machine learning research. They coordinate researchers around shared research problems and serve as a measure of progress towards shared goals. Despite the foundational role of benchmarking practices in this field, relatively little attention has been paid to the dynamics of benchmark dataset use and reuse, within or across machine learning subcommunities.

In this paper, we dig into these dynamics. We study how dataset usage patterns differ across machine learning subcommunities and across time from 2015-2020. We find increasing concentration on fewer and fewer datasets within task communities, significant adoption of datasets from other tasks, and concentration across the field on datasets that have been introduced by researchers situated within a small number of elite institutions.

Our results have implications for scientific evaluation, AI ethics, and equity/access within the field.

"We’ve developed an aesthetics of complexity: the sense that a good system is a complex one, that you should prefer a SPA over a web page, a distributed system over a simple one, a service over a config file, the idea if you aren’t on the latest technology you’re wasting your time, and potentially damaging your career."

Same news but from OpenAI: https://news.ycombinator.com/item?id=34488811

Abstract:

Probability measures that are constrained to the sphere form an important class of statistical models and are used, for example, in modeling directional data or shapes. Therefore, and as building block methodology, efficient sampling of distributions on the sphere is highly appreciated. We propose a shrinkage based and an idealized geodesic slice sampling Markov chain, designed to generate approximate samples from distributions on the sphere. In particular, the shrinkage based algorithm works in any dimension, is straight-forward to implement and has no algorithmic parameters such that no tuning is necessary. Apart from the verification of reversibility we show under weak regularity conditions on the target distribution that geodesic slice sampling is uniformly geometrically ergodic, i.e., uniform exponential convergence to the target is proven.

Abstract:

The acceptance of automated driving is under the potential threat of motion sickness. It hinders the passengers' willingness to perform secondary activities.

In order to mitigate motion sickness in automated vehicles, we propose an optimization-based motion planning algorithm that minimizes the distribution of acceleration energy within the frequency range that is found to be the most nauseogenic. The algorithm is formulated into integral and receding-horizon variants and compared with a commonly used alternative approach aiming to minimize accelerations in general.

The proposed approach can reduce frequency-weighted acceleration by up to 11.3% compared with not considering the frequency sensitivity for the price of reduced overall acceleration comfort.

Our simulation studies also reveal a loss of performance by the receding-horizon approach over the integral approach when varying the preview time and nominal sampling time. The computation time of the receding-horizon planner is around or below the real-time threshold when using a longer sampling time but without causing significant performance loss.

We also present the results of experiments conducted to measure the performance of human drivers on a public road section that the simulated scenario is actually based on. The proposed method can achieve a 19\% improvement in general acceleration comfort or a 32% reduction in squared motion sickness dose value over the best-performing participant.

The results demonstrate considerable potential for improving motion comfort and mitigating motion sickness using our approach in automated vehicles.

Abstract:

The aviation literature gives relatively little guidance to practitioners about the specifics of architecting systems for safety, particularly the impact of architecture on allocating safety requirements, or the relative ease of system assurance resulting from system or subsystem level architectural choices.

As an exemplar, this paper considers common architectural patterns used within traditional aviation systems and explores their safety and safety assurance implications when applied in the context of integrating artificial intelligence (AI) and machine learning (ML) based functionality.

Considering safety as an architectural property, we discuss both the allocation of safety requirements and the architectural trade-offs involved early in the design lifecycle. This approach could be extended to other assured properties, similar to safety, such as security.

We conclude with a discussion of the safety considerations that emerge in the context of candidate architectural patterns that have been proposed in the recent literature for enabling autonomy capabilities by integrating AI and ML. A recommendation is made for the generation of a property-driven architectural pattern catalogue.

Abstract:

In many ways, graphs are the main modality of data we receive from nature. This is due to the fact that most of the patterns we see, both in natural and artificial systems, are elegantly representable using the language of graph structures. Prominent examples include molecules (represented as graphs of atoms and bonds), social networks and transportation networks. This potential has already been seen by key scientific and industrial groups, with already-impacted application areas including traffic forecasting, drug discovery, social network analysis and recommender systems. Further, some of the most successful domains of application for machine learning in previous years -- images, text and speech processing -- can be seen as special cases of graph representation learning, and consequently there has been significant exchange of information between these areas.

The main aim of this short survey is to enable the reader to assimilate the key concepts in the area, and position graph representation learning in a proper context with related fields.

Abstract:

Federated Learning (FL) is a popular distributed machine learning paradigm that enables jointly training a global model without sharing clients' data. However, its repetitive server-client communication gives room for backdoor attacks with aim to mislead the global model into a targeted misprediction when a specific trigger pattern is presented. In response to such backdoor threats on federated learning, various defense measures have been proposed.

In this paper, we study whether the current defense mechanisms truly neutralize the backdoor threats from federated learning in a practical setting by proposing a new federated backdoor attack method for possible countermeasures. Different from traditional training (on triggered data) and rescaling (the malicious client model) based backdoor injection, the proposed backdoor attack framework (1) directly modifies (a small proportion of) local model weights to inject the backdoor trigger via sign flips; (2) jointly optimize the trigger pattern with the client model, thus is more persistent and stealthy for circumventing existing defenses.

In a case study, we examine the strength and weaknesses of recent federated backdoor defenses from three major categories and provide suggestions to the practitioners when training federated models in practice.

Abstract:

Nowadays, many social media platforms are centered around content creators (CC). On these platforms, the tie formation process depends on two factors: (a) the exposure of users to CCs (decided by, e.g., a recommender system), and (b) the following decision-making process of users.

Recent research studies underlined the importance of content quality by showing that under exploratory recommendation strategies, the network eventually converges to a state where the higher the quality of the CC, the higher their expected number of followers.

In this paper, we extend prior work by (a) looking beyond averages to assess the fairness of the process and (b) investigating the importance of exploratory recommendations for achieving fair outcomes. Using an analytical approach, we show that non-exploratory recommendations converge fast but usually lead to unfair outcomes. Moreover, even with exploration, we are only guaranteed fair outcomes for the highest (and lowest) quality CCs.