Source: Xinzhiyuan

A new round of AI battles is about to begin!

On Monday, OpenAI will launch an online live broadcast, officially announce the GPT-4 upgrade, and even a super “AI assistant” is waiting for us.

Alexis Conneau, the “head of audio AGI research” at OpenAI, has changed the homepage background and is on the same frequency as Altman - we'll see Magic next week.

OpenAI research scientist Bowen Cheng even said that this is much cooler than GPT-5.

All of this suggests that the real “Her” is about to debut.

Under pressure from OpenAI, Google will announce new model developments at the I/O conference the next day. Rumor has it that it will also release a personal digital assistant called “Pixie”, sponsored by Gemini.

Immediately after that, Microsoft will hold a Build developer conference on the 21st. It is likely that it will integrate OpenAI's latest capabilities into its product line, and may even reveal MAI-1, the latest 500 billion parameter self-developed model.

There is also the much-anticipated Apple WWDC conference, which will release the iOS 18 system with integrated generative AI capabilities and embed ChatGPT into the iPhone.

The series of major announcements and repeated bombardment simply gave other companies no chance to catch their breath.

Some netizens asked, “Is Apple abandoning its' AJAX 'artificial intelligence system and making every effort to cooperate with OpenAI? Or is OpenAI just a stopgap until their AI capabilities catch up”?

Apple insider Gurman summed up Apple's artificial intelligence strategy:

- Device-side LLM (self-developed)

- LLM in the cloud (self-developed)

- Chatbots (maybe OpenAI or Google)

Apple doesn't plan to develop a chatbot on its own, but it realizes that the market is in demand for it, so it will acquire this technology from outside. This strategy is similar to their approach in the search field.

Obviously, the current situation is that OpenAI is tied up with Microsoft and even Apple through AI cooperation, leaving Google alone.

I don't know, who will win and who will lose in this wave of AI battles for supremacy?

ChatGPT can be called, and Zhou always broadcast the news again

The focus of the entire network is still OpenAI.

“What will they publish” is a topic that only continues to grow in popularity, and few people discuss the Google I/O conference.

Regarding Monday's release forecast, netizen Ananay made another discovery:

ChatGPT may have the ability to make calls

In fact, this function can be seen from the following code, from keywords such as making a call or refusing a call.

Additionally, OpenAI has deployed WebRTC servers to implement this functionality, and these servers have also recently been configured.

At first, netizens thought that this was probably because OpenAI deployed a WebRTC server in voice-only mode, but now it seems that this is not the case.

Because, this feature is provided by Livekit. (This is a solution that can provide real-time audio and video communication)

A netizen below commented, does this mean that ChatGPT can take the initiative to call me without me having to make a call first?

He raised this question because in the movie Her, Samantha, an artificial intelligence assistant, took the initiative to call the male protagonist to tell him something.

Imagine how amazing it would be for the ChatGPT Assistant to take the initiative to call you, remind you, or check user habits.

However, Ananay said this requires users to choose to allow this feature.

Indigo, the co-founder of Hallid.ai, also made a comprehensive prediction/trend conjecture.

According to Indigo, the new version of GPT-4 should be divided into multiple versions according to the scale of the parameters.

Yesterday, some netizens speculated that gpt4-lite, gpt4-auto, and gpt4-lite-auto versions might be released.

And the GPT2-Chatbot that appeared at the LMSYS Arena a few days ago is probably a new lightweight GPT-4 version. Moreover, this means that the mission of GPT-3.5 is coming to an end. The latest lightweight version may be free to use, and the API price will drop drastically.

As for the “magic” Altman refers to, it is probably the upgraded GPT-4 - GPT4-Auto, which has the ability to independently perform agent tasks, has stronger memory, and stronger planning ability.

Of course, the “AI Assistant” also brought Her into reality.

Yesterday, OpenAI video generation research scientist Will DePue posted a logo showing the advent of a singularity, which is probably hinting at something.

Google plays a ring, or launches Pixie, an AI assistant

At this critical moment of rivalry with OpenAI and Microsoft, Google made it clear that the content released at this conference was all about AI.

According to Google's official website, this year's I/O conference will be held on May 14 at 1 p.m. EST.

It is speculated that Google will integrate generative AI into the search engine to allow users to conduct conversational searches.

In addition, Google has also been testing new search features, such as AI conversation exercises for English learners and generating virtual try on images while shopping.

More than just a search engine, more Google apps will also integrate AI functions more deeply, such as helping users find suitable restaurants, shopping centers, and electric vehicle charging stations in Google Maps.

What should I do when I call customer service and the transfer takes too long?

The new AI feature tested by Google can even help you automatically wait for transfers until someone answers and then notify you.

Apart from all kinds of apps, the operating system must not be left behind.

The Android 15 developer preview was released last month, and Google will further introduce the new features at the I/O conference and may add deeper Gemini integration.

Currently, in the Android system, generative AI functions are mainly driven by Gemini Nano and used in various software functions.

For example, Magice Compose can provide response suggestions in apps such as Google Messages, and Cinematic Wallpaper uses machine learning to help users customize screen wallpapers.

Can you imagine what more personalized user experiences Android will bring with further AI participation? Like a smarter phone home screen, lock screen, and notification bar?

At last year's I/O conference, we saw Gemini, a big language model that competes with ChatGPT. Will there be a new model this year?

In addition to the new version of Gemini, you can probably look forward to the big image and video models launched by Google.

Netizens on Reddit broke the news that 3 models in Google's inventory have been tested but have not yet been released to the public. It is estimated that they will be unveiled at the 2024 I/O conference.

The three models are Imagen 3, an image generation model, and Juno and Miro, two models that can optimize and complete images.

It is said that Miro will also have a video generation function.

Furthermore, Google may release a new version of the AI assistant “Pixie” at I/O this year, which may replace the original Google Assistant, a similar product.

Pixie is driven by the language model Gemini and is installed on the Pixel, a hardware device developed by Google itself. We don't know if it will be open to other third-party devices.

However, we should not see the updated version of the Pixel product at this I/O conference. Google has recently released a new version of the Pixel 8a, and it is already open to users to pre-order purchases.

The Pixel 9 and the folding Pixel 9 Pro Fold are expected to be released this fall.

Apple clings to the lifesaver

Meanwhile, in the face of the menacing impact of AI voice assistants from OpenAI and Google, netizens shouted at Apple:

There's not much time left for Apple!

Although it is already reported that OpenAI and Apple are about to finalize a cooperation agreement to enable ChatGPT to be installed in iPhones and provide new generative AI capabilities for this year's iOS system.

But Apple isn't ready to give up its own Siri.

Recently, The New York Times reported that Apple will upgrade and restructure Siri to deal with other chatbot competitors.

And that decision has already been made.

At the beginning of 2023, Apple executives Craig Federighi and John GiannAndrea felt a deep crisis after spending weeks testing the ever-popular new OpenAI chatbot ChatGPT.

They think the advent of generative artificial intelligence has made Siri seem outdated and backward.

As the first virtual assistant in every iPhone launched by Apple in 2011, Siri has always been limited to satisfying individual requests and cannot keep up with conversations initiated by users.

For example, someone first asked about the weather in San Francisco and then said, “How's New York?” From time to time, Siri often misunderstands users' questions.

But ChatGPT knows that what users want is an answer to the latter question.

After realising that new technology had surpassed Siri, the tech giant underwent its most significant restructuring in more than a decade.

Apple is determined to catch up with the artificial intelligence competition in the technology industry. Using generative artificial intelligence as a special benchmark project within the company, it organizes employees around a once-in-ten year plan.

Siri Super Evolution

According to three Apple insiders, Apple will release the improved Siri at the annual developer conference on June 10 this year.

The underlying technology in the new version includes new generative artificial intelligence that will allow Siri to chat with users rather than answer questions one at a time.

It also makes Siri more conversational and more versatile.

The Siri update is one of Apple's leading moves to fully embrace generative AI.

To support its new Siri features, memory has also been added to this year's iPhone.

Additionally, Apple also discussed the possibility of partnering with several companies, including Google, CoHere, and OpenAI to obtain the right to use AI models that support chatbots.

On the other hand, Apple executives are also worried that emerging AI technology will replace iOS as the main operating system in the future, threatening Apple's dominant position in the global smartphone market.

Furthermore, this new technology may also enable an ecosystem centered on AI applications (AI agents).

This could weaken Apple's App Store, which has sales of around $24 billion a year.

However, what Apple is more worried about is that if it cannot develop its own AI system, the iPhone may become a “dumb phone” in comparison with other advanced technologies and lose its market.

The iPhone currently accounts for 85% of global smartphone profits and has generated more than $200 billion in sales.

This loss can be expected to be immeasurable, and Apple cannot accept it.

The urgency of this crisis prompted Apple to cancel another major investment --

A $10 billion autonomous vehicle project and dispatched hundreds of engineers to AI development work.

Furthermore, Apple will continue to maintain consistent device process tools and explore the creation of servers powered by iPhone and Mac processors.

According to insiders, Apple's upgrade to Siri is not about making it compete with ChatGPT for content generation such as poetry creation, but rather for Siri to focus on handling its original tasks:

This includes setting alarms, creating calendar reminders, adding items to shopping lists, and summarizing text messages.

Apple plans to claim that the upgraded Siri will provide more private services and are more cost-effective than competing companies' artificial intelligence.

Because Siri processes requests on the iPhone, this avoids data leaks in the cloud and the costs of cloud computing.

However, Apple also faces the risk of small artificial intelligence systems installed on iPhones:

The study found that smaller artificial intelligence systems may be more likely to hallucinate than larger systems.

Tom Gruber, co-founder of Siri, said:

“Siri's goal has always been to create a conversational interface that understands language and context, but it's a challenge.

As technology changes, we should be able to do better. As long as you don't try to solve everything in the same way, you can avoid a lot of difficulties.”

Apple has many advantages in the field of artificial intelligence, including more than 2 billion devices in use around the world, a leading semiconductor team, etc.

They can support Apple in promoting AI products, and support AI tasks that require a large number of chips, including facial recognition.

Can Apple reverse the situation in a month

However, in the past ten years, Apple has never formulated a comprehensive artificial intelligence strategy, and Siri has not had major upgrades since its launch.

At the same time, its limitations as a voice assistant also diminish the appeal of the company's smart speaker HomePod, because it is unable to perform simple tasks stably, such as responding to song playback requests.

John Burkey, who founded the generative artificial intelligence platform Brighten.ai after working on the Siri team for two years, said:

“Since its inception, the Siri team has not received the same level of attention and resources as other teams within Apple.

However, Apple's different departments are often independent, and information sharing is limited.

But the truth is that AI needs to be integrated into products to be successful.”

In addition, Apple also has considerable resistance in recruiting and retaining leading artificial intelligence talents.

Due to Apple's confidentiality, few research results have published papers or attended conferences, which is an almost unbearable disadvantage for scientists.

In recent months, Apple has slightly adjusted its consistent strategy to increase the number of artificial intelligence papers published, but industry researchers still question the quality of the papers, believing that they are Apple's marketing hype.

But for some budding and ambitious researchers, joining Apple as a leading member of the project is an important reason they chose Apple.

Although Apple has adjusted its development strategy and absorbed quite a bit of fresh blood.

However, in this huge and dazzling battle for AI voice assistants, it is still unknown whether Apple can reverse its disadvantage at the June developer conference.

What will the future shape of AI voice assistants look like, and how will they affect our lives?

The answer to this question is getting closer and closer to us.

edit/lambor