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.

The rise of AI in language translation has led to questions about the future of human translators. While some fear AI will render human translators obsolete, the sources suggest a more nuanced picture. AI translation tools offer undeniable advantages in speed, efficiency, and cost-effectiveness, particularly for large-scale and straightforward translations. However, AI still falls short when it comes to cultural nuance, handling ambiguity, and translating specialized or creative content.

The consensus among experts is that AI will augment, rather than replace, human translators. AI tools can assist translators by generating initial drafts, suggesting terminology, and maintaining consistency. This collaboration frees human translators to focus on the more complex aspects of language, ensuring accuracy, cultural appropriateness, and stylistic finesse.

The sources highlight several areas where human expertise remains crucial:

Conveying cultural and emotional nuances: Humans excel at understanding idioms, humor, and culturally specific references, which AI often misinterprets.

Accuracy in specialized fields: Technical, legal, and medical translations demand precise language and domain expertise, areas where AI is still developing.

Creative language use: Translating literature, marketing materials, and poetry requires creativity and stylistic sensitivity, something AI currently lacks.

Ultimately, the future of translation likely involves a symbiotic relationship between AI and human expertise. AI handles the routine tasks, while humans provide the nuanced understanding and creative touch, resulting in more efficient and effective communication across languages.

ChatGPT, the popular AI chatbot, played a notable role in the 2024 US election, with millions of users turning to it for information. OpenAI, the company behind ChatGPT, implemented safety measures to prevent the spread of misinformation, including: Banning impersonation of candidates or governments. Discouraging misrepresentation of voting procedures and voter suppression.

Digitally watermarking AI-generated images created with DALL-E. OpenAI also collaborated with the National Association of Secretaries of State to: Provide accurate answers to election-related queries. Direct users to CanIVote.org, a non-partisan voting information hub. These efforts resulted in over a million ChatGPT responses directing users to CanIVote.org and over 2 million responses pointing users to reputable news sources like the Associated Press and Reuters for election results.

Additionally, ChatGPT rejected over 250,000 requests for deepfakes of prominent political figures. Despite these measures, the Bipartisan Policy Center expressed concerns about potential misinformation, highlighting the limitations of AI chatbots in providing complete and consistent information.

They urged users to verify information from ChatGPT with reliable sources like government websites or local election boards. The 2024 election marked the first time voters could use ChatGPT for election information, signalling a potential shift in how people engage with political processes.

While OpenAI's efforts demonstrated a commitment to mitigating misinformation, the Bipartisan Policy Center's cautionary note suggests that there's room for improvement before future elections.

Google's Gemini AI model, while impressive in many areas, has demonstrated significant shortcomings when tasked with stock market prediction. To evaluate Gemini's capabilities, the model was asked to identify the top 10 gainers and losers from the NASDAQ-100 Index over a 100-day period.

Gemini's responses were riddled with errors, including:

Listing fewer than 10 stocks.

Repeating stocks within a list.

Including stocks not belonging to the NASDAQ-100.

Naming the same stock as both a gainer and a loser.

Producing three inconsistent lists for each question.

While Gemini's stock selections occasionally outperformed the NASDAQ-100 Index, the overall results were inconsistent and puzzling. For example, stocks predicted to be the biggest losers often showed gains, highlighting Gemini's difficulty in distinguishing between winners and losers.

Gemini's performance on similar tests using the CAC 40 and AEX indices further underscored its limitations in stock prediction.

Gemini is not a reliable stock advisor due to its technical errors and questionable financial results.

Interestingly, Gemini seems to have acknowledged its shortcomings. When asked the same questions today, it provides a detailed explanation of its limitations and why it cannot provide definitive answers.

In a response to the research, Gemini acknowledges the complexity of financial markets and its own limitations as a language model. It emphasizes its primary function as processing and generating text and its unsuitability for real-time financial analysis. Gemini also expresses confidence in its continuous improvement and the potential for future iterations to handle complex tasks like financial forecasting.

Midjourney, a popular AI image generator, is rolling out a new feature allowing users to edit external images uploaded from their computers or online. This expansion moves Midjourney beyond editing AI-generated images and positions it as a competitor to traditional photo editing software like Photoshop. The editor allows users to move, resize, erase, and retexture images using text prompts.

However, this new functionality raises concerns about copyright infringement, as it enables users to modify existing images without permission. Midjourney is already facing a class action lawsuit from artists alleging copyright infringement in training its AI models. The lawsuit has entered the discovery stage, which means the artists' lawyers can examine Midjourney's training datasets and potentially reveal information about the company's practices.

One of the lead plaintiffs, Kelly McKernan, expresses hope that the lawsuit will force AI companies to use licensed content and compensate artists. She also hopes the case will establish legal protection for artistic styles under the Lanham Act.

Midjourney acknowledges the need for careful moderation to prevent misuse of the new editing tool. The company is initially restricting access to users with annual memberships, long-term subscribers, or those who have generated a significant number of images.

The long-term impact of AI image editing tools on the art community remains to be seen. While some artists fear these tools may devalue their work and lead to job losses, others see potential for collaboration and new creative opportunities.

Chegg, an online education company known for textbook rentals and homework help, has been significantly impacted by the emergence of AI chatbots like ChatGPT. Chegg's stock price has plummeted by 99% since its peak in 2021, leading to a loss of $14.5 billion in value and half a million paid subscribers. This decline is largely attributed to ChatGPT offering a free and readily available alternative to Chegg's paid services.

Students have been abandoning their Chegg subscriptions, which cost around $20 per month, in favor of the free and instant assistance provided by ChatGPT. A survey conducted by Needham, an investment bank, revealed a decline in Chegg usage among college students, with only 30% intending to use it, compared to 62% planning to use ChatGPT.

Chegg's struggles highlight the disruptive potential of AI in the education technology sector. Other companies like Coursera, Udemy, Course Hero, and Quizlet are also experiencing challenges due to AI's emergence. The accessibility, breadth of knowledge, and free nature of AI chatbots pose a significant threat to these companies' business models.

While Chegg had internal discussions about AI and its potential impact, they underestimated the speed of its development and consumer adoption. Despite warnings from employees and external advisors, Chegg executives believed an experience like ChatGPT wouldn't be possible until 2025.

Chegg is now attempting to adapt by developing its own AI-powered study assistant, CheggMate. However, the company faces an uphill battle in convincing customers and investors of its value in a market dominated by ChatGPT. CheggMate's success remains uncertain, particularly given the company's historical reliance on acquisitions rather than in-house product development. Furthermore, concerns about cheating and the ethical implications of AI in education persist.

Automated Insights (AI) is an American tech company specializing in natural language generation (NLG) software, which turns large datasets into readable narratives. AI's main product is Wordsmith, a platform that transforms structured data into written content, allowing businesses to automate content creation, such as reports, articles, and other written materials.

One of AI's most notable clients is the Associated Press (AP), which uses Wordsmith to automate financial and sports reporting. This partnership began in 2014, and the AP's use of Wordsmith has enabled them to increase their output of earnings reports significantly, from producing around 300 earnings stories per quarter to nearly 4,000. This automation has freed up journalists' time, allowing them to concentrate on more in-depth reporting and investigative journalism. The AP has also used AI to generate previews for NCAA Division I men's basketball games, covering over 5,000 regular-season games.

Wordsmith uses NLG and machine learning to convert data into readable narratives. This process involves using data from sources like STATS for sports and Zacks Investment Research for financial reporting. AI applies natural language processing and advanced algorithms to these datasets to generate human-readable content.

While AI can be a valuable tool for journalists, it's important to remember that it shouldn't replace human creativity and judgment. The most effective journalism will always combine human and artificial intelligence.