This summary explores OpenAI's ChatGPT Search and its effect on news publishers, based on analysis and testing by the Tow Center. Despite OpenAI’s claims of collaborating with the news industry, their search tool often misrepresents and misattributes content. For example, ChatGPT frequently cites sources incorrectly, including attributing quotes to the wrong publication or providing incorrect dates and URLs.

The Tow Center's testing revealed that ChatGPT rarely admits when it can't locate a source. Instead, it often fabricates information, making it hard for users to judge the validity of the information provided.

OpenAI allows publishers to block ChatGPT’s access to their content via a “robots.txt” file. However, the Tow Center found that ChatGPT still referenced blocked content, sometimes citing plagiarised copies found on other websites. This raises concerns about OpenAI’s commitment to accuracy and attribution.

The study ultimately highlights the limited control publishers have over how their content appears in ChatGPT Search. Even publishers with licensing agreements with OpenAI experienced misattribution and misrepresentation of their content. The Tow Center concludes that OpenAI needs to address these issues to ensure accurate representation and citation of news content.

A recent study reveals that large language models (Large Language Models Reflect the Ideology of their Creators), like ChatGPT and Google’s Gemini, exhibit inherent ideological biases reflecting their creators' viewpoints.

The research, employing a novel methodology, analysed LLMs' descriptions of historical figures to identify these biases, finding significant differences between Western and non-Western models, and even variations based on the language used.

These findings challenge the notion of AI neutrality, highlighting the influence of training data and refinement processes on an LLM's output. The study's authors emphasise the need for transparency regarding these biases and suggest alternative regulatory approaches to mitigate potential risks of political manipulation and societal polarisation. The implications for information access and the future of AI regulation are significant.

xAI, Elon Musk's artificial intelligence startup, is rapidly expanding, securing significant funding and building a massive supercomputer in Memphis.

The company aims to create advanced AI models, such as the Grok chatbot, challenging established players like OpenAI. xAI's ambitious goals include developing AI capable of complex mathematical reasoning and ultimately, understanding the universe.

However, its rapid growth has sparked controversy surrounding its environmental impact, funding sources, and potential monopolistic practices.

The company's close ties to Musk's other ventures and its close relationship with the incoming Trump administration also raise concerns.

Today we discuss a growing concern: AI's potential to exacerbate "dark patterns," manipulative design tactics employed on websites and apps to influence user behaviour. These patterns, present on a vast majority of popular platforms, trick users into actions they might not otherwise take, such as unwanted subscriptions or purchases.

While AI personalisation can enhance user experience, it becomes problematic when it exploits vulnerabilities or personal data to push products or services. The real danger lies in generative AI's ability to amplify these dark patterns, replicating them from the vast datasets it learns from.

One study revealed that AI-powered language models like ChatGPT consistently integrated dark patterns into website designs, even when given neutral prompts. This raises ethical and legal questions about AI's role in normalising manipulative designs.

Although regulation and consumer awareness are crucial to combatting this issue, the sources also highlight businesses embracing ethical design practices. By prioritising transparency and user choice, companies can build trust and long-term customer loyalty, ultimately creating a digital landscape where fairness is the norm. The sources argue that eliminating dark patterns benefits both users and businesses in the long run.

OpenAI is encountering performance limitations with its new Orion large language model (LLM), which is expected to succeed GPT-4. While Orion is showing better performance in some areas compared to its predecessors, it's not demonstrating the same level of improvement seen between GPT-3 and GPT-4. The improvements are notably smaller, particularly in tasks such as coding.

The limited availability of high-quality training data is a major contributing factor to these challenges.. As LLMs have consumed a significant amount of publicly available data, finding new sources of good-quality training data is becoming more difficult and expensive. OpenAI is exploring several strategies to address this challenge. These strategies include:

Creating a “foundations team” to investigate methods for improving LLMs with declining data availability.

Training Orion on synthetic data generated by other AI models.

Optimizing LLMs at a stage after the initial training phase..

Reliance on synthetic data, however, poses its own risks, such as model collapse, where future models merely replicate the abilities of their predecessors. OpenAI is aware of these risks and is working on strategies to mitigate them.

The cost of developing and running these advanced models is also rising due to the increasing computational resources required for training. This raises concerns about the economic viability of future LLM development.

Despite these challenges, OpenAI remains committed to innovation and is exploring new approaches to enhance AI models. The company is also considering a potential shift in the naming scheme for its next-generation AI model, moving away from the "ChatGPT" naming convention.

The use of AI tools like ChatGPT in education has sparked debate about its potential to facilitate cheating. While some students may use these tools for legitimate purposes, there is growing concern about their use for generating essays and completing assignments, raising concerns about academic integrity. A survey revealed that a significant number of students consider using AI for assignments and exams as cheating, and a substantial portion have already done so.

The ease with which AI can produce credible academic work has led to increased distrust among educators and a rise in disciplinary actions against students suspected of using AI for cheating. Efforts to combat this issue include developing AI detection tools and revising assignments to require more original thought. However, the effectiveness of these measures is debated, with some arguing that AI detection technology is lagging behind AI content generation.

Some experts believe that the focus should shift from policing cheating to adapting teaching methods to incorporate AI ethically. This includes emphasizing writing as a process of intellectual development and teaching students to use AI responsibly. Ultimately, the challenge lies in finding a balance between harnessing the potential benefits of AI in education while upholding academic integrity.

Liquid AI, an MIT spin-off, has launched its first series of generative AI models called Liquid Foundation Models (LFMs). These models are built on a fundamentally new architecture, based on liquid neural networks (LNNs), that differs from the transformer architecture currently underpinning most generative AI applications.

Instead of transformers, LFMs use "computational units deeply rooted in the theory of dynamical systems, signal processing, and numerical linear algebra". This allows them to be more adaptable and efficient, processing up to 1 million tokens while keeping memory usage to a minimum.

LFMs come in three sizes:

LFM 1.3B: Ideal for highly resource-constrained environments.

LFM 3B: Optimised for edge deployment.

LFM 40B: A Mixture-of-Experts (MoE) model designed for tackling more complex tasks.

These models have already shown superior performance compared to other transformer-based models of comparable size, such as Meta's Llama 3.1-8B and Microsoft's Phi-3.5 3.8B. LFM-1.3B, for example, outperforms Meta's Llama 3.2-1.2B and Microsoft’s Phi-1.5 on several benchmarks, including the Massive Multitask Language Understanding (MMLU) benchmark.

One of the key advantages of LFMs is their memory efficiency. They have a smaller memory footprint compared to transformer architectures, especially for long inputs. LFM-3B requires only 16 GB of memory compared to the 48 GB required by Meta's Llama-3.2-3B.

LFMs are also highly effective in utilizing their context length. They can process longer sequences on the same hardware due to their efficient input compression.

While not open source, users can access LFMs through Liquid's inference playground, Lambda Chat, or Perplexity AI. Liquid AI is also optimising its models for deployment on various hardware, including those from NVIDIA, AMD, Apple, Qualcomm, and Cerebras.

Artificial intelligence (AI) is rapidly changing the video game industry, offering potential benefits and raising serious concerns. While some celebrate its potential to revolutionize game development, others, including many game developers, fear its impact on their livelihoods.

AI’s potential benefits for game developers include:

Reducing development costs and time: AI can automate time-consuming tasks, such as creating 3D environments, populating game worlds with assets, and testing gameplay.

Enhancing game quality: AI can help developers analyze data to improve game performance, identify and fix bugs, and create more realistic graphics and animations.

Personalizing the gaming experience: AI can tailor storylines, adjust difficulty levels, and create dynamic environments based on player preferences.

However, concerns exist about AI's potential negative impact on game developers:

Job displacement: As AI becomes more sophisticated, it could replace human artists, writers, and level designers, particularly those performing routine tasks.

Deskilling and job degradation: Some fear that artists, rather than creating original work, will be relegated to fixing AI-generated content.

Ethical concerns: The use of AI raises questions about copyright, ownership, and potential biases in algorithms.

The future of work in the gaming industry likely involves a hybrid model, with AI tools augmenting human creativity and skill.

The key is to ensure that AI is used to enhance the gaming experience rather than replace human ingenuity. This will require the industry to address ethical concerns, upskill its workforce, and foster collaboration between humans and AI.