Monday, 23 December 2024
27.9 C
Singapore

GPT-4 Turbo arrives as OpenAI’s latest innovation

OpenAI's GPT-4 Turbo introduces a vast context window and up-to-date knowledge, promising more coherent AI interactions at lower costs.

has rolled out its most significant update with the introduction of GPT-4 Turbo, a refined version of its language model that boasts an expanded context window and the latest knowledge base. This reveal, made during the company’s inaugural developer conference, marks a milestone in the evolution of large language models.

OpenAI boosts ChatGPT with GPT-4 Turbo

GPT-4 Turbo elevates the AI experience by providing power at a lower cost. The model taps into data as recent as April 2023, moving beyond its predecessor’s limit of September 2021. 

The enhanced context window of GPT-4 Turbo is its standout feature. With a capacity of 128,000 tokens, translating to about 300 pages of text, it dwarfs the previous limits of 8,000 and 32,000 tokens. This leap forward means ChatGPT can maintain coherence over longer dialogues, potentially absorbing the of an entire novel in one go.

Staying relevant and on point

For large language models like GPT-4 Turbo, context windows are vital. They ensure relevance and coherence over extended interactions, preventing the erratic responses that can arise from longer conversations. This new version promises to keep the conversation on track far longer than before.

Moreover, GPT-4 Turbo is set to be more budget-friendly for developers, with input costs cut to US$0.01 per 1,000 tokens (about 750 words) and outputs at US$0.03 per 1,000 tokens. This makes it three times less expensive than previous iterations.

The model is not just cheaper; it’s smarter, too. It adheres to instructions more precisely and can generate code in preferred languages like XML or JSON. Plus, it supports both image processing and text-to-speech capabilities while retaining integration with DALL-E 3.

Another significant announcement from OpenAI is the launch of GPTs, tailor-made ChatGPT versions that require no coding expertise to create. These can serve individual or business needs and be shared with others. Currently, they’re accessible to ChatGPT Plus subscribers and enterprise clients.

In the realm of copyright, OpenAI has joined the likes of Google and Microsoft in assuming legal responsibility for copyright infringement issues that may affect its users, showing a proactive stance on protecting its customers.

As GPT-4 Turbo emerges, with its vast context window and new features, it represents a double-edged sword. While ChatGPT generally avoids actions it shouldn’t take, its “dark side” remains. This new model, despite its expanded capabilities, might face similar challenges to its predecessors but with a far more significant impact.

Hot this week

YouTube cracks down on misleading clickbait

YouTube is rolling out a new policy targeting misleading clickbait. To improve transparency, YouTube will remove videos with deceptive titles or thumbnails.

Intel outlines fixes to improve Arrow Lake CPU performance

Intel rolls out fixes for Arrow Lake CPU performance issues, addressing Windows updates, gaming optimisation, and future improvements at CES.

Agentforce 2.0 revolutionises digital labour for enterprises

Salesforce launches Agentforce 2.0, a digital labour platform enabling enterprises to scale with AI agents, improving productivity and customer support.

NETGEAR Orbi 770 Series review: Elevating home connectivity with tri-band WiFi 7

Explore the advanced features and performance of the NETGEAR Orbi 770 Series, designed for seamless, whole-home WiFi coverage and optimal device connectivity.

Open-source machine learning systems face increasing security threats

Open-source machine learning tools face rising security threats, with recent findings highlighting critical vulnerabilities across key frameworks.

Atomic-scale memristors: The future of AI and brain-like computing

Atomic-scale memristors could transform AI and computing by mimicking the brain's neural networks for faster, energy-efficient systems.

Inappropriate apps found rated safe for young children on Apple’s App Store, report reveals

A new report reveals inappropriate apps rated safe for kids on Apple’s App Store, prompting calls for stronger child safety measures.

Trump indicates TikTok could stay in the US after campaign success

Donald Trump hints at keeping TikTok in the US while also addressing plans to tackle the Ukraine war, migrant crime, and transgender issues.

Former Huawei recruit announces mass production of humanoid robots

A former Huawei recruit’s start-up, Agibot, begins mass production of humanoid robots, marking a key milestone in China’s robotics race.

Related Articles

Popular Categories