Artificial Intelligence (AI) is the ultimate intern in the technological landscape, demanding continuous oversight and careful management. While its capabilities are vast, the spectre of misuse and unintended consequences casts a formidable shadow. Yet, recent steps in AI development offer a glimmer of optimism, signalling a movement towards more ethical practices.
As we reflect on the initial quarter of the year, a rich tapestry of advancements, regulations, and debates shapes the AI narrative, shedding light on both its achievements and hurdles. From ethical initiatives to groundbreaking model releases and regulatory landmarks, Q1 unveils a blend of innovation and scrutiny, emphasising the pivotal role AI holds in our ever-evolving world.
What's New with OpenAI
- Making Steps to Ethical AI Practices
The startup takes measures to ensure ethical AI practices by investing in the training of its models. The company made licensing deals with two major European publishers, the French paper Le Monde and the Spanish media conglomerate Promotora de Informaciones SA, to bring French and Spanish-language news to ChatGPT and help train their models. These deals are part of OpenAI’s plan to work with media companies instead of fighting with them about how they use news in their AI tools. It’s good to see AI companies are paying to train their models. This helps ensure that the AI’s results are accurate, reliable, and trustworthy in the long run. - Sora: OpenAI’s Instant Video Maker
This February OpenAI introduced Sora, its new AI tool capable of creating instant videos based on written instructions. Named after the Japanese word for “sky,” Sora stands out among previous Text-to-Video (T2V) tools for its sophistication. With Sora, a short text description can be transformed into a detailed, high-definition video clip lasting up to a minute. - Introducing Figure 01: OpenAI’s Humanoid Robot
OpenAI and the robotics startup Figure AI have developed a humanoid robot called Figure 01 equipped with advanced capabilities derived from a new visual language model. This robot utilises OpenAI’s multimodal model, VLM, to VLM, to comprehendimages and texts, enabling “speech-to-speech” analysis. Unlike other OpenAI technologies that rely on written prompts, Figure 01 generates responses based on voice conversations. Its code evolves with each interaction through Figure’s neural network, promising continuous improvement as it engages with more users.
The European Union's AI Act
The European Union’s AI Act represents a significant milestone in regulating artificial intelligence (AI) within the EU. It marks the first comprehensive set of rules outlining the permissible use of AI systems. Approved on 13 March 2024, this Act categorizes AI systems into four groups based on their societal risks. For instance, high-risk applications such as self-driving cars are subject to stringent regulations before deployment in the EU. Conversely, AI systems with lower risks face comparatively fewer restrictions.
The focus now shifts towards ensuring compliance among businesses and institutions, with additional AI-related legislation tailored for workplaces underway. The AI Act is set to take effect in May 2025, with stricter regulations for high-risk systems slated to be implemented three years later. National agencies will play a pivotal role in monitoring compliance across the EU to uphold the Act’s provisions.
Anthropic’s States Claude 3 Beats ChatGPT4
Anthropic’s claims that their latest language model, Claude 3 has surged ahead of competitors like ChatGPT and Google’s Gemini, setting new industry standards in performance and capability. The company highlights that Claude 3 has not only surpassed its predecessors but has also achieved “near-human” proficiency in various tasks. Anthropic attributes this success to rigorous testing and development, culminating in three distinct chatbot variants: Haiku, Sonnet, and Opus. In benchmark tests, Opus emerged as a frontrunner, outperforming GPT-4 in various tasks, while Sonnet showcased remarkable speed and intelligence. Moreover, Claude 3’s enhanced visual processing capabilities mark a significant advancement, ensuring a nuanced understanding of user requests and enhancing productivity.
Latest from Google's AI Model Gemini
- Google launches Gemini 1.5
Google has introduced its latest AI model, Gemini 1.5, featuring an “experimental” one million token context window. This new feature allows Gemini 1.5 to understand context and meaning in extremely long text passages, far surpassing previous AI systems like Claude 2.1 and GPT-4 Turbo. According to Google researchers, Gemini 1.5 Pro achieves near-perfect recall on long-context tasks across various areas. Its innovative Mixture-of-Experts (MoE) architecture enhances efficiency by selectively activating the most relevant expert pathways in its neural network.
- Gemini’s inaccurate and biased image generation
On the other side, Google’s Gemini model has faced criticism for producing historically inaccurate and racially skewed images, sparking renewed concerns about bias in AI systems. Social media users have shared examples of Gemini generating pictures depicting racially diverse Nazis, black medieval English kings, and other improbable scenarios. While there’s recognition of the importance of diversity in AI-generated content, some argue that Google’s response has been too extreme. Experts emphasise the centralisation of AI models under a few major corporations and advocate for the development of open-source AI models to encourage diversity and address bias on a broader scale.
Elon Musk launched Grok
Named after the term “understanding” from Douglas Adams’ Hitchhiker’s Guide series, Grok is positioned as a more open and humorous alternative to other models like OpenAI’s ChatGPT. This move aligns with Musk’s advocacy against censorship and his recent legal battle with OpenAI over open sourcing AI technology.
Elon Musk’s startup xAI has released its large language model Grok as open-source software, fulfilling Musk’s promise to accelerate AI development and adoption. The 314 billion parameter model is now freely available under an Apache 2.0 license, allowing anyone to access, modify, and distribute it. xAI shared the base model weights and network architecture of Grok-1, their advanced language model trained from scratch. Open sourcing Grok empowers developers, companies, and enthusiasts worldwide to leverage its capabilities.
Wrap up
As we navigate the intricate web of advancements and debates, one thing remains clear: AI’s trajectory is deeply intertwined with our collective journey towards responsible innovation and societal impact. Through concerted efforts in regulation, collaboration, and ethical development, we can harness the transformative power of AI while mitigating its potential risks, ensuring a future where technology serves humanity’s best interests.
Leave a Comment