- Lumiera
- Posts
- 🔆 GPT4 OH?! The TLDR of OpenAI's latest launch.
🔆 GPT4 OH?! The TLDR of OpenAI's latest launch.
The latest on ChatGPT. Also: Is your PIN code a safe choice and systems to combat school shootings.
Lumi Era
May 16, 2024
🗞️ Issue 18 // ⏱️ Read Time: 4 min
Hello 👋
In this week's newsletter
What we’re talking about: OpenAI announced GPT-4o, their fastest and most affordable flagship model, as part of their recent updates.
How it’s relevant: OpenAI is one of the leading AI companies, developing AI models that are used globally. Other similar companies, such as Google, are going to be looking to outshine each other.
Why it matters: The enhanced capabilities of the AI systems is going to affect human-AI interaction. It indicates in which direction we are headed with AI development.
Short news snippets about other things in the AI world
🖥️ Google DeepMind, in collaboration with Isomorphic Labs, has introduced AlphaFold 3, a groundbreaking artificial intelligence (AI) model capable of predicting the structures and interactions of all life's molecules with unparalleled accuracy.
🌍 The software company Iterate.ai is offering a free artificial intelligence system that can help security detect weapons at schools to combat school shootings.
⚖️ TikTok is testing AI-generated search results, a feature that presents AI-generated content at the top of the search results page for certain queries.
📱 The most common pin codes. Phone, credit cards and door locks have something in common: They use a pin for access. How predictable is your chosen PIN number? 3.4 million data points visualised from several data breaches give insight into which four digits are the most used.
The TLDR of the GPT-4o Announcement
Omni is the o in GPT4o. It means all and refers to the many different kinds of media used: text, audio, video and image. This multimodality leads to more natural human-computer interactions, an intention that is explicitly stated in the official announcement of the launch. The Digital Assistant market as we know it is gone: Amazon’s Alexa and iPhone’s Siri won’t be staying around for much longer.
What is new?
🌍📚 Available to everyone, also non-paying users. GPT-4 is significantly better than GPT-3.5. This is an important step, as it’s making GenAI more accessible and essentially democratising knowledge.
🗣️💬 Real-time voice conversations: GPT-4o enables real-time voice interactions, allowing humans to engage in conversations with the model with very little delay. It makes the responsiveness of the AI model feel more natural.
👁️📸 Vision capabilities: These new capabilities enable GPT-4o to analyse images, screenshots, and desktop screens to provide relevant information. It can answer questions based on this visual content.
🧠📈 Memory breakthrough: Memory functionality allows GPT-4o to maintain context and coherence throughout interactions. The experience for a human using the model may be that the model remembers more.
🌐🗣️ Support for over 50 languages: A full list is not yet available, but a few examples stated so far are Korean, Italian and English.
📊🔍 Advanced data analysis tools: These tools can analyze data, create charts, and upload files. It makes it possible to extract insights, generate charts, and obtain answers from various types of data.
🎙️🗨️ Voice Mode and Voice Interaction: Voice commands and responses are now part of the interaction experience, making it very conversational.
📷🖼️ Ability to chat about and generate images from photos and screenshots: GPT-4o can analyse visual content from photos and screenshots - you can engage in discussions and receive information related to these visual elements.
What is the same?
GPT-4o is still just a large language model, which comes with its limitations:
📅✂️ Cut off date still exists, for GPT-4o it is October 2023. This means that the data that the LLM bases the responses on is at least 7 months old.
🤖⚠️ It hallucinates, does not learn from experiences like humans do, does not forget like humans do, and has potential for errors.
A few observations
It’s worth noting that this launch was made the day before the Google I/O announcements. Anthropic also announced that Claude is now available in Europe. A strategic move by OpenAI to gain attention? Team Lumiera thinks so.
What will this mean for human-AI interaction and collaboration? The comparison to the HER, the movie where a lonely writer develops a relationship with an operating system designed to meet his every need, is everywhere on the internet.
What are some other smart observations?
Ed Newton-Rex, CEO of Fairly Trained pushes for fair training data.
Sinead Bovell, Founder of WAYE comments on the risks with human-AI relationships.
Ian Pay, Head of Data Analytics & Tech at ICAEW thinks accuracy should be prioritised.
Zoe Kleinman, Techonology reporter at BBC points out the environmental impact
Until next time.
On behalf of Team Lumiera
Lumiera has gathered the brightest people from the technology and policy sectors to give you top-quality advice so you can navigate the new AI Era.
Follow the carefully curated Lumiera podcast playlist to stay informed and challenged on all things AI.
What did you think of today's newsletter? |
