The only thing constant in this world is Change. The internet is abuzz with all the recent turmoils in the AI world and its only going to get even more intense. Disruptions can be classified into 5 types irrespective of the industry they impact, and with so much changing in the world of AI I think its high time we understand these and see how disruptions manifest when applied to the world AI.
1) Disruptions to the Tool:
This disruption is probably the most visible and most frequent and most hated. This is also due to the fact that we are closest to our tools to get our different tasks done. So any disruptions that dethrone the incumbent tool and replaces it with a new one is bound to strike up emotions. While these kinds of disruptions are frequent and feel personal, they are transient with today's disruptor becoming tomorrow's disrupted. On the bright side, already established Industries are not impacted due to these disruptions and the loss and gains are always limited to individual companies or products.
Disruptive Manifestations in AI
- ChatGPT to Deepseek - We recently had such an episode when DeepSeek dethroned the incumbent ChatGPT. The shock was so massive that it sent the stock markets tumbling. They re-moulded our core beliefs on whats possible and not possible with AI. And people either loved it or in the case of Wall street they hated it! ChatGPT Disrupted!
- MS Excel to Copilot - Another tool level disruption that is being talked about is Microsoft Excel being replaced by AI driven Microsoft 365 Copilot agents. This upgrade eliminates the need to use Excel to analyze data, plot graphs and draw insights. All of this and more can now be easily done by just typing prompts into a Copilot agent that is linked with your dataset. MS Excel Disrupted!
2) Disruptions to the Technique:
These disruptions are characterized by their ability to make an existing technique simpler and more efficient. Unlike the tool based disruption these disruptions are often welcomed and less feared by the end user. Thats also because the deeper workings of the improvements are often hidden from view sparing the end users the burden of an abrupt change. These disruptions also do not have the power to disrupt established incumbent industries.
Disruptive Manifestations in AI
- RAG to Agentic RAG - Microsoft recently updated their Microsoft 365 Co-Pilot chat application to allow users to easily build and launch their own Agents. One of the most talked use case for these agents is enabling users to ground their newly built Agents on specific documents for retrieving information and answering questions related to those documents. This is exactly what RAG was all about. Albeit in the days of the good old AI this process of building RAG systems was time consuming and needed expert technical knowledge. With this new UI for AI, as Microsoft calls it, building cumbersome RAG systems are now a thing of the past. Companies investing in AI to build in-house RAG systems now have the ability to get their system up and running in no time with absolutely technical expertise on RAG concepts. RAG Technique Disrupted!
3) Disruptions to the Architecture:
Architecture based disruptions run deep. They alter the very fabric of the system they have an impact on. These disruptions often target an inherent inefficiency of the old system by devising new and effective ways to overcome limitations. These disruptions are comparatively fewer in number when compared to the earlier 2 types of disruptions, but such disruptions could lead to innovations in new Tools and Techniques that could in turn result in spawning of new industries and products.
Disruptive Manifestations in AI
- RNN to Transformers - Recurrent Neural Networks or RNNs was the dominant AI architecture till 2017 when it was dethroned by the introduction of a new Transformer based architecture. RNNs were good at processing input sequences one word at a time, but they struggled with long-range dependencies. This meant that they could not relate back to words and phrases that were used earlier in a conversation. This created problems in Natural Language Processing (NLP) tasks where it was important to maintain memory of previously used words and sentences to enable a meaningful conversation. This was solved by the Transformer architecture by introducing a new Self-attention mechanism that processed the entire sequence of words in parallel. This change in architecture was responsible for the popularization of Chatbots as we know today. This was also responsible for introducing a new industry based on LLM chatbots. RNNs Architecture Disrupted!
- Transformers to SSMs - The Transformer architecture is now falling out of favor due to their high energy consumption and slow speed. Solving for this inefficiency are State Space Models (SSMs) like the Mamba that run faster and cheaper. Its only a matter of time before the Mamba too takes over from the Transformer. Transformer Architecture soon to be Disrupted!
4) Disruptions to the Ecosystem:
An ecosystem based disruption alters the entire environment of the incumbent process. You not only change the way you do stuff, but rather you alter the entire place where you get the stuff done. The primary motive driving these disruptions is to rise the bar of what's possible, this happens when the incumbent technology has reached its practical limits and is no longer possible for accomplishing anything more. Such disruptions are relatively rare due to its dependence on multiple variables and the effort needed to reinvent a new wheel. But these disruptions are so powerful that they generate new industries, tools and techniques that are far beyond current day imaginations.
Disruptive Manifestations in AI
- Transistor Chips to Quantum Chips - In Dec 2024 Google introduced to the world their latest Quantum Chip "Willow". According to Google, Willow could solve in under five minutes, what would take the fastest current day supercomputers 10 septillion years to solve. This is soon to be followed by Microsoft's Majorana 1 (an article on this coming soon). The prospects of a rising computing demand for AI coupled with the limitations of the current day Transistor based chips was what prompted Google for this disruption. Moore's law had predicted Silicone based Transistor chips to max out in the next few years, so the only viable option was to move out of the Silicon based Transistor ecosystem into the Quantum realm where the bars could be set higher, much much higher. Transistor Chip based Ecosystem soon to be Disrupted!
5) Disruptions to the Enduser:
Enduser disruptions often found bordering on Human Psychology and Philosophy, tampering with the very motive driving our behavior. Obviously such disruptions are very rare and take a long time to materialize. They have the ability to make the enduser relook at not just the solution or the way we solve a problem, but instead question the very problem itself.
Disruptive Manifestations in AI
- Reactive to Proactive - All along humanity has been obsessed with predicting the future. But one thing we needed for making accurate predictions was fast and high quality data from the past. Hence most of our modern day technology has been striving to optimize communications to ensure our data was relevant and clean. Business and empires were built on the ability to transmit fast and accurate information. With the advancements in modern day computing infrastructure this is now no longer of concern. Our technology has now enabled us to transmit data from the past with remarkable speed, high accuracy and low costs. With the massive advancements in AI and computing, our target now is to predict the future. We have started to become Proactive instead of being Reactive. We use all the AI Tools, Techniques, Architectures and Ecosystems to do just one thing, Predict the Future! This disruption in our efforts to predict the future from capturing the past is already happening. Businesses capable of seeing into the future scanning for new opportunities will be the ones that survive. We will move from being backward looking to forward seeking and our future will be defined with this forward seeking attitude! And this is the biggest disruption of the new AI enabled Era! Human Reactive attitude currently in Disruption!
This blog was partly inspired by Clayton M Christensen's amazing book The Innovators Dilemma, that introduced me to the Dynamics of Disruption.
#AI #Agentic #Artificial #Intelligence #Agents #Implementation #Algorithm #Data #Inputs #Outputs #Solutions #Stages #Disruptions #Disruption #Deepseek #Copilot #Excel #Microsoft #Quantum #Computing #Silicone #Transistors #RNN #Transformers #SMM #Mamba #RAG #Reactive #Proactive #Tools #techniques #Architecture #Ecosystem #Enduser
PS: As an certified AI consultant with 19 years experience in Marketing, Analytics & AI, I can help SMBs explore the power of AI to benefit their business. Whether you're looking for AI implementation, data analysis, or customer insights, I can tailor solutions to your specific needs. Feel free to leave a comment or send me a direct message to discuss your AI goals.
Analytics| Business Planning| Business Intelligence| Digital Transformation| Data Enablement| Product Management| Data Management| Data Modelling| Revenue & Pipeline Management| Storyteller| Stakeholder Management
1 周great post Sriharsha Ganjam, thank you for sharing.
Busy Building AI Solutions | Google Cloud Expert | M.Sc. Electrical Engineering
2 周Great article!
Ex-CMD NFL, Noida, Uttar Pradesh, India. Chairman RFCL
1 个月Thanks for sharing Sriharsha Ganjam . Very insightful analysis of the entire ecosystem of AI??