The big news over recent weeks has been Meta integrating Meta AI into the search fields of their mobile apps. The company upgraded its chatbot to Llama 3 and it is now running it in the search bars of Facebook, Messenger, Instagram and WhatsApp. Alongside this, the company launched other new features, such as faster image generation and access to web search results.

Why is this significant?

For those of us interested enough in AI to write or read newsletters about it, we see AI everywhere. However, the vast majority of people either don’t know or don’t seem to care too much. Pew Research reported that as of March, only 23% of US adults had tried ChatGPT, and 34% had never even heard of it.

Bob Knorpp, in a great piece for Datos Research on the impact of AI on search, shows that “the typical (ChatGPT) user will investigate the tool, try it a few times and then seemingly loses interest”.

This is where Meta’s move is a potential game-changer. Facebook has around 3 billion monthly active users, and around 2 billion each on Instagram and WhatsApp. All these users have now been onboarded onto AI whether they like it or not. 

Compare these billions to ChatGPT’s 180 million users and the scale becomes apparent. Not only this, but Meta AI’s introduction has been far from subtle. Microsoft’s Copilot and Google’s Gemini have been gently introduced as optional tools and features. Meta has put it front and center by replacing the search function with the Ask Meta AI, exposing billions to AI recommendations, image and content creation.

Will Meta eat itself?

This begs the question of whether we, and AI, are ready. Trust in social media has been constantly degrading: from fake news and disinformation to AI influencers and deep fakes. Pushing AI generation at such scale, even before Meta has found a successful solution to identifying and tagging AI content, has the potential to drown the platforms in even more AI content than we currently witness.

Additionally, Llama-3, while a very capable model, still has many issues. Users are already reporting that Meta AI cannot be switched off, but if you ask it, it will lie and tell you that you can. It has also started posting in groups, in one example claiming to be the parent of a gifted, disabled child.

This has been a big gamble by Meta to force AI into regular use by billions. Time will tell in showing its impact and user-acceptance, but this is a watershed moment in bringing AI to the masses.

The hot thing in AI is…GPT2?

The AI community has been buzzing over the last couple of days with the mysterious emergence and quick disappearance of gpt2-chatbot.

Released without documentation on LMSYS Arena, users were quickly impressed by it’s capabilities, solving a Math Olympiad problem and appearing to be better at complex code than Claude Opus or GPT4.

Speculation quickly grew about what this model was, whether it was OpenAI’s gpt-2 with improved training or a secret test for GPT5.

We may never know as the mysterious model has now disappeared with no word or comment from its creator.

ChatGPT gets memory

OpenAI has rolled out a new feature for ChatGPT Plus subscribers called Memory. The AI chatbot can store details that you share in conversations and refers to this information during future chats.

Announced in February, ChatGPT’s Memory feature had been available only to a small group of users but is now available to Plus subscribers, except for those in Europe and Korea.

Read the tutorial down below to find out how to enable the feature, make it remember and also make it forget.

Did a friend forward this e-mail to you?

IE₊ SUPPORT INTERESTING ENGINEERING
_{Invest In Science And Engineering}

Enjoy exclusive access to the forefront of AI content, highlighting trends and news that shape the future. Join a community passionate about AI, delve into the latest AI breakthroughs, and be informed with our AI-focused weekly premium newsletters. With IE+, AI reporting goes beyond the ordinary - and it is Ad-Free.

NEWS

• Adobe’s AI-powered VideoGigaGAN makes videos 8x clearer

  Adobe’s latest AI tool gives special attention to fine details like skin textures and even individual strands of hair.
  

• TUM professor turns drones into dancers using ChatGPT, here’s how

  Robotics researchers from TUM just used ChatGPT to program a swarm of drones to perform a choreography in the air.
  

• Sanctuary AI’s latest 7th-generation Phoenix learns tasks in just 24hrs

  The new model features greatly improved software and hardware, making it the fastest and smartest to date.

MUST READ

Meet Vidu: China’s answer to OpenAI’s text-to-video generator Sora

Chinese startup Shengshu Technology and Tsinghua University have officially unveiled China’s answer to OpenAI’s Sora, Vidu. The AI-powered text-to-video app can generate 16-second clips at 1080p resolution with a single click.

While considerably shorter than Sora’s 60-second video capability, Vidu is the best China currently offers. The new text-to-video software was unveiled on the weekend at the Zhongguancun Forum in Beijing.

Before you search for Vidu, there is an existing product of the same name (but capitalized) that is, in fact, a tool for sales teams.

“Unfortunately for us, they chose the name “Vidu” for their AI model. We’ve been using the name VIDU for our product since 2021,” VIDU explains.

AI PICTURE OF THE WEEK

TUTORIAL

How to Use ChatGPT’s Memory

Using ChatGPT’s Memory is as simple as just chatting with the bot. As you use ChatGPT it learns from the conversations. You may see a ‘Memory updated’ notification as you interact. Tap on the notification to see what was remembered.

You can force information into the memory by directly asking ChatGPT to ‘Remember this…” or similar.

You can also change what it remembers by prompting it with “forget everything you remember about…” or “adjust my location to...”.

To see everything ChatGPT remembers about you, click on your username, go to Settings and Personalization, and then tap on the Manage button at the bottom of the screen.

You can select the trash can icon next to any memory you’d like to remove. You can also select ‘Clear ChatGPT’s Memory’, which will let you start over with a clean slate.

Memory is enabled automatically, but you can opt-out by opening Settings, then Personalization, and then toggling the Memory option.

For privacy, OpenAI’s FAQ reads, “We may use content that you provide to ChatGPT, including memories, to improve our models for everyone.” You can disable this through Data Controls in settings and OpenAI state they “won't train on content from ChatGPT Team and Enterprise customers.”

TOOLS OF THE WEEK

Amazon Q: Amazon have launched their AI-powered assistant generally available for developers and businesses as well as released free courses on using the AI assistant and a new Amazon Q capability in preview. Amazon Q is designed to act as a secure, AI assistant for businesses, capable of leveraging internal data and performing technical tasks such as coding, debugging, multi-step planning, and reasoning to help developers optimize their workflows, as well as help employees get answers to questions across business data.

Legal Robot: Legal Robot automatically translates complex and confusing “legalese” into straightforward language. Useful both for people wanting to make sure they understand legal documents and for legal professionals to ensure that their contracts and documents are written in terms that anyone can understand.

Lalal.ai: Lalal.ai automates audio source separation, extracting elements such as vocals, music, or even specific instrumental tracks like drumbeats or basslines from any audio or video content.