February 2025

Perplexity Just Made AI Research Crazy Cheap—what That Means for the Industry

,

Perplexity's new tool, Deep Research, drastically reduces AI research costs, challenging high enterprise subscription models. It offers similar capabilities to major providers like OpenAI and Google at a fraction of the cost, potentially reshaping AI pricing and accessibility. Perplexity provides five free queries daily, with pro plans at $20/month for 500 queries, making advanced AI available to small businesses and researchers. This shift may force competitors to justify their higher prices as AI investments continue to rise.

https://venturebeat.com/ai/perplexity-just-made-ai-research-crazy-cheap-what-that-means-for-the-industry/

Detecting AI Agent Use & Abuse

, ,

TLDR: AI agents now mimic human users online, complicating detection of both legitimate and malicious activity. Traditional methods like CAPTCHAs and IP blocking are ineffective against sophisticated agents. Applications must enhance observability to differentiate between human and AI traffic. Consumer sites like Reddit and YouTube block some AI agents, but many sites lack adequate detection capabilities. Adapting detection strategies with machine learning is essential as AI agents evolve, balancing user experience with security.

https://stytch.com/blog/detecting-ai-agent-use-abuse/

Introducing Perplexity Deep Research

,

Perplexity Deep Research launched, automating in-depth research for users. It performs extensive searches and analysis, generating comprehensive reports in minutes. It's free for all, with unlimited queries for pro subscribers. It's available on the Web now, with apps for iOS, Android, and Mac coming soon. Deep Research excels in various fields, achieving high accuracy on major benchmarks and allowing easy report sharing.

https://www.perplexity.ai/hub/blog/introducing-perplexity-deep-research

Ollama

,

Apple's Ollama enables running local large language models (LLMs) on Macs, similar to Docker for AI. It ensures privacy, eliminates costs, reduces latency, and offers control and reliability. Ollama's HTTP API allows easy app integration using Swift, with capabilities like text generation, chat interactions, and embeddings. Local processing means documents remain private. The tool promotes immediate innovation, contrasting with delays from Apple Intelligence. Examples like Nominate.app showcase Ollama’s practical applications. In short, Ollama facilitates immediate AI-powered app development without reliance on external models.

https://nshipster.com/ollama/

Which Economic Tasks Are Performed With AI?

,

AI tasks primarily involve software development and writing, with significant usage across the economy. About 36% of jobs integrate AI into tasks, focusing on augmenting human abilities (57%) versus automating tasks (43%). Younger, educated, higher-income workers in sectors like customer service and IT are the main users of generative AI, which raises concerns over potential atrophy in critical skills as reliance grows. The risk is that AI's convenience may lead to shallow thinking and intellectual complacency, threatening original thought in sectors reliant on cognitive processes.

https://marginalrevolution.com/marginalrevolution/2025/02/which-economic-tasks-are-performed-with-ai.html

Adobe’s Sora-rivaling AI Video Generator Is Now Available for Everyone

, ,

Adobe has launched its AI video generator, Generate Video, in public beta, allowing users to create five-second video clips at 1080p. It includes features for generating videos from text and images, with customizable options for styles and camera angles. The tool outputs video in 1080p at 24fps, but has longer generation times compared to competitors like OpenAI's Sora. Adobe's Firefly web app also integrates with Creative Cloud and offers AI-generated assets safely for commercial use. New subscription plans provide credits for using the tool, with additional paid features expected.

https://www.theverge.com/news/610876/adobe-generate-video-ai-public-beta-available

AI History: Key Milestones That Shaped Artificial Intelligence

,

AI has transitioned from theoretical concepts to a vital component of technology and daily life. This article examines AI's historical milestones, from early achievements in the 1950s-60s, through periods of stagnation and resurgence, to current advancements like deep learning and generative AI. It discusses how AI systems mimic human intelligence, applications across industries, and future goals like achieving AGI and ASI. Today, AI influences communication, problem-solving, and creativity, raising ethical considerations for responsible development.

https://www.grammarly.com/blog/ai/ai-history/

Large Behavior Models Surpass Large Language Models To Create AI That Walks And Talks

,

Large behavior models (LBMs) are advancing AI by integrating generative AI with behavioral training, enabling robots to learn tasks through observation and interaction. Unlike traditional large language models (LLMs) that primarily use language, LBMs also consider behaviors, improving AI's ability to mimic tasks like cooking by observing human actions. As LBMs develop, they require careful ethical considerations and adaptability to ensure safe and effective AI applications in real-world environments.

https://www.forbes.com/sites/lanceeliot/2024/11/10/large-behavior-models-surpass-large-language-models-to-create-ai-that-walks-and-talks/

Scroll to Top