Sam Altman, the CEO of OpenAI, is optimistic about the deployment of artificial general intelligence. He believes that by 2025, machines will think and reason like humans.
As Elon Musk launched Grok-3 recently, many wonder if the AGI is already here. Some claim this is true, while others believe it is an overstretch of this utopian optimism. Whatever the case, no one can deny that with tech giants like OpenAI, xAI, Google, and now DeepSeek investing in research, we are headed to the AGI revolution.
This journey towards achieving general intelligence is marked by notable developments and predictions. So, let’s get a quick rundown of the AGI timeline from AlphaGo to Grok-3, and the challenges ahead.
In this article
Part 1. What Is AGI
AGI, or artificial general intelligence, is theoretical AI research that intends to mimic human cognition. Simply put, AGI refers to the hypothetical AI machine with the intelligence of a human brain that tends to self-teach and solve problems it was never trained for.
What makes AGI different from artificial intelligence is its generalization ability, which transfers knowledge and skills to a domain different from what it was educated for. Additionally, it has a vast collection of knowledge (facts, social norms, etc.), depending on which it can reason and make decisions.
That said, no true AGI system exists today, though some applications, like self-driving cars and conversational chatbots, possess some general intelligence qualities. AI can still not replicate the level of success possible with human intelligence. However, if ever developed, AGI would exceed human capacities, given its ability to process huge data sets faster.
According to AI researchers, here are some core AGI capabilities.
- Creativity: Be able to read and understand human-generated codes and improve them.
- Sensory perception: Excel at color recognition and perceive the depth in static images.
- Motor skills: It involves a lot of imaginative perception.
- Natural language understanding: Comprehend the meaning of context-dependent human language.
- Navigation: Direct movements through physical spaces like a global positioning system.
Part 2. AGI Timeline: From AlphaGo to Grok3
The beginning of AGI
The transition from weak AI to strong AI results from decades of advancements. But what truly marks this shift was Geoffrey Hinton's work on neural networks, which could loosely be defined as a system that processes information like a human brain.
What started in the 1970s took shape in 2012 when Hinton and his two graduate students published their research at the ImageNet competition. Their research mostly explored neural networks and deep learning mechanisms capable of processing big datasets and making accurate decisions.
Let’s explore the AGI timeline from here onwards.
below. 
below. 
2014 – DeepMind’s AlphaGo
The first practical event that marked the transition to strong AI was the development of AlphaGo. The concept of this AI program was taken from an ancient game of Go; an incredibly difficult play for a computer to comprehend with the vast number of possible movements.
Developed by the AI research lab DeepMind, this game used reinforcement learning. It was trained using a combination of advanced search algorithms and neural networks.
2016 - Microsoft Tay
Microsoft launched its original Twitter bot named Tay in March 2016. However, it was deployed within hours for posting offensive and hurtful tweets. Microsoft claimed that the trolls caused this, as the bot was designed to make replies by interacting with people on Twitter.
Additionally, DeepMind’s AlphaGo beat one of the best global players, Lee Sedol, in 2016. (Source: BBC News)
2017 - Google’s The Transformer
A team of Google researchers proposed a new network architecture called the Transformers. This architecture was based on recurrent neural networks and the self-attention mechanism.
2018 - OpenAI launched GPT
The researchers and engineers at OpenAI, notably Alec Radford, published his work on the first GPT, Generative Pre-Training. Their study was focused on how a large language model can gain knowledge based on pre-training on a diverse and big dataset.
2019 - GPT-2
One year later, OpenAI launched the improved GPT-2 language model. Unlike the previous version, this advanced language model was trained on more than nine mission documents. However, this development raised concerns about the potential misuse of AI-produced content.
2020- GPT-3
From 2020 onwards, the AI revolution has been all thanks to developments in generative AI, the ability of AI to produce images, texts, and videos. These advancements are seen as a building block toward the broader ambition of achieving general intelligence.
Following the release of GPT-2 in 2019, OpenAI developed its new model, GPT-3 LLM, in 2020. It consists of 175 billion parameters and can generate human-like text. During this time, Nvidia also announced its beta version of Omniverse, which would produce 3D models in the physical world.
2021- DALL.E
OpenAI introduced DALL.E. This AI system was based on the GPT-3 modified model and could generate images from text commands.
2022 - OpenAI launched ChatGPT
In November 2022, OpenAI released ChatGPT-3.5, which would gain over one million users within five days of its launch. This would access the web data up to 2021 and answer user’s queries in a conversational style.
Some other notable developments in strong AI during 2022 include,
- A startup business, Stability, released Stable Diffusion, a generative AI technology that creates images from users' text.
- The Midjourney Version 1 and 2 were also launched in 2022.
- Additionally, DALLE-2 also entered the beta phase.
- Google DeepMind announced AlphaTensor. It could discover provably correct algorithms for complex mathematical problems.
2023 – Significant Developments in GenAI
- Google released its generative chatbot called Bard.
- Microsoft integrated ChatGPT technology into Bing, a free-to-use GenAI for Microsoft users.
- OpenAI released a beta version of the ChatGPT browser extension. Improved from a previous model, this LLM processes both images and texts as prompts. However, at the time, this tool was only accessible to Plus subscribers.
- Midjourney versions five and six were released.
- DeepSeek released DeepSeek coder, the first open-source model for specific code-related tasks. The company also released its first version of the multipurpose LLM model.
- xAI announced Grok in November 2023, marketed as a conversational chatbot. According to xAI, this bot would use wit and humor “to answer spicy questions rejected by most other AI systems”.
- Elon Musk and Steve Wonzaik, along with 31000 other tech leaders, wrote a letter urging a six-month pause due to the AI arms race. According to the letter, companies were launching AI systems that couldn’t be understood or controlled.
Geoffrey Hinton, the Godfather of AI, quit Google due to the dangers of AI development. He, along with other tech leaders, issued a warning about the risks AI presents to human existence, which were as massive as the nuclear war.
2024 - xAI Released Grok-2
After announcing Grok-1.5 in May, xAI launched Grok-2 in August 2023, a language model with state-of-the-art reasoning abilities in text and vision understanding. The Grok-2 and Grok-2 mini were made accessible for all Grok users on the X platform. Like other AI systems, these models excel at chatting, coding, and reasoning.
2025 - DeepSeek R1, o3, and Grok-3 Released
DeepSeek: On January 20, 2025, DeepSeek released its R1 large language model with an open-source license. Within days of its launch, the DeepSeek mobile app with a chatbot interface reached the top of Apple’s App Store, outranking ChatGPT apps.
Unlike Open AI’s conversational models, DeepSeek’s training involves less time and limited AI accelerators at a fraction of the cost. DeepSeek aims to achieve AGI with significant progress in AI reasoning capabilities. Some innovations developed in the R1 model include reinforcement learning, distillation, reward engineering, and emerging behavior networks.
O3: Additionally, OpenAI launched its o3 mini medium and o3 mini high with an advanced level of reasoning and exceptional STEM capabilities. According to the company, it is the most cost-efficient language model, trained with a massive scale of reinforcement learning.
Grok 3: Following this, on February 12, 2025, xAI launched the Grok 3 chatbot. According to Musk, this AI system would outperform ChatGPT and DeepSeek and perform better across math, coding, and science-related benchmarks. This system is trained on 200,000 Nvidia H100 GPUs, 15 times more powerful than Grok-2.
The company also announced a new tool called DeepSearch and called it the next-generation search engine. The Grok-3 and DeepSearch will be accessible exclusively to Premium Plus members on the X platform.
Part 3: Ethical Implications of General Intelligence
As modern machines aim to exceed human intelligence, they offer opportunities and ethical challenges. Training a machine to align with human values and societal norms seems the hardest. Let’s understand some essential ethical implications of AGI if it is developed in the future.
Societal Tensions and Instability
A critical aspect of AGI's ethical implications remains societal changes and instability. The fact that AGI’s power would be in the hands of a few influential entities could exacerbate the existing social and economic injustices.
One major concern in this regard is job displacement. Many jobs will become obsolete as AGI systems become powerful enough to perform multiple tasks across industries. This can ultimately result in economic disruptions.
Moral Dilemmas
Like the previous AI systems, the research and deployment of artificial general intelligence pose moral dilemmas, mainly the ethical treatment of the technology. If these systems reach the level of anticipated consciousness, this will raise questions about human rights and moral obligations we might possess towards them.
The use of AGI is especially concerning in decision-making processes across industries, including healthcare, the military, and law enforcement. Given the complexity of human ethics and biases, it would be challenging for the AGI to make unbiased and fair decisions.
Existential Threat
If ever developed, the general and super-intelligence would exceed humans' cognitive abilities. This concept challenges the core assumption of human exceptionalism, making us secondary intelligent beings. Additionally, AI becoming more self-aware and intelligent than us raises questions of human consciousness and personhood.
Part 4: How to Explain Concepts with Timelines using EdrawMInd?
Timelines are an excellent way to demonstrate the evolution of a concept or an event. They give your viewer a graphical representation of chronological history that is easy to visualize. Project managers, educators, and researchers use these diagrams to break down the sequential order of a happening.
However, making them is a whole new story. Thankfully, digital software has made it easy to plot the sequence of a timeline. One such tool is EdrawMind, which lets you plan projects and draw historical events while promoting team communication.
Here is how you can use this timeline maker for project planning.
Step 1
Download the template from above and start editing in the canvas space. Or, open the EdrawMind desktop version and visit the Template Gallery to explore similar timeline examples.
Step 2
Once on the canvas, start by adjusting the timeline's structure. To do this, visit the Start tab and expand Topics, Sub-Topics, etc.
Step 3
Now, you can add information. Switch to the Outline mode from the top-left corner for a linear view. It lets you get a clear idea of the data so you can make modifications accordingly.
Step 4
The next step is adding visual cues. It helps retain your viewer’s attention better. So, open the left-side customization panel and place clipart in your timeline.
Step 5
It’s almost done. All you need to do is modify the design details. To do this, visit the Page Layout tab and change the theme, theme color, and background image accordingly.
Step 6
Ta-da, your timeline is ready to be shared. Visit the File menu from the top-left corner, click Export, and choose a desired format. Alternatively, click Share from the quick-access toolbar to invite your team members for real-time edits and instant feedback.
AGI: To Be or Not To Be?
There you have it: the AGI timeline from AlphaGo to Grok-3! The time to achieve this technology is uncertain, with varying views from experts.
That said, given the current AI landscape and AGI predictions, it is only wise to be ready for technology to become smarter if not wholly surpass human intelligence. While we are moving toward an uncertain path, it is equally important to focus on the ethical, governance, and moral challenges it brings.
EdrawMind
AppsEdrawMind Online
