Imagine this: The year is 2002. You are fortunate enough to obtain a cutting-edge smartphone that enables you to message anyone on the planet. Changing one’s life, huh? During the early 2000s, firms such as Ericsson, Nokia, and BlackBerry dominated the mobile phone market. When the iPhone first came out in 2007, everything changed and the former market leaders were wiped out.
Trackvital | lookyourlooks | about-local | theoneseehub | quincyoffers | event-allstars | gallerytokoku | riocapitals | server-garden
The iPhone revolution shows us that long-term winners are not always the ones that innovated first during a tech boom cycle. As a matter of fact, they rarely do. This is an important thing for all founders and VCs to think about, especially because the AI hype cycle is still fluctuating and early-stage generative AI businesses are being valued at exorbitant sums.
Why all the excitement about AI?
An enormous wave of momentum began in the field of generation AI with the release of OpenAI’s ChatGPT. Since then, the product has been used by 92% of Fortune 500 businesses, and almost every significant big tech player has released a version of it. Simultaneously, numerous “wrapper” firms surfaced with products that expand upon ChatGPT’s framework.
The propensity of humans to exaggerate change in the short term compared to the long term is one aspect that undoubtedly contributed to the accumulation. Reversals in projections on the replacement of jobs by AI have already occurred. For instance, the World Economic Forum projected in 2020 that by 2025, AI would displace 85 million jobs globally. However, according to their most recent estimate, AI is predicted to generate a net number of jobs.
Although there is no denying AI will disrupt the workplace, as timescales are accelerated, the hype bubble expands. Once more, prior hype cycles highlight the need of avoiding making such assertions. Another instance of this is the early 2010s, when significant advancements in computer vision and speech recognition were made possible by important neural network research.
The hyperbole that usually fuels technological hype cycles was summed up in a 2013 Popular Science article that said, “We should probably just accept the fact that we are that much closer to the sentient-robot conquest.” This is not to downplay the importance of the deep learning achievements of 2012; rather, it is to argue that we may learn from the past to make sense of the current AI frenzy. Fourteen years later, the robots are still not in control, but the everyday gadgets we use are becoming more efficient and effective.
How to judge whether the buzz around an AI startup is justified
When deciding where to lay your bets, there are a few things to take into account given how frothy the AI market is right now. Similar to any gold rush, people naturally want to find the picks and shovels so they can explore and build things, or, to put it another way, make horizontal tools and infrastructural solutions.
At the same time, one must be aware that the rate of change is a significant distinction between this platform shift and previous ones. Both well-established tech companies and upstarts are revolutionizing their technology platforms at the same time, and large technology platform suppliers are also demonstrating astonishing flexibility in their adaptation. When compared to the early days of the cloud build, this translates into a considerably faster evolution of the build with gen AI stacks.
In the event that data and computation represent the currency of innovation in generation AI, we must consider how startups compare structurally to more established tech incumbents and how much access to compute they have (many foundation model companies have also raised enormous sums of money to buy that access).
Higher up the stack, the potential for applications appears to be extremely large; but, considering where we are in the hype cycle, the legal environment, the developments in cybersecurity posture, and the dependability of AI outputs are important gating considerations that must be addressed before commercial adoption at scale can occur.
Finally, the performance of foundation models is a result of their pre-training on internet-scale datasets. Building models in more industry-specific domains will require the ability to compile sizable, high-quality datasets, which is a necessary step towards realizing the benefits of AI. It is becoming more and more obvious that the models themselves are not the primary differentiator; rather, it is the quantity and quality of data used for training.
Maintaining awareness of regulations
Global regulatory organizations have taken notice of the excitement and vast potential for transformation that generative artificial intelligence (gen AI) and large language models (LLMs) provide. Startups must prepare for potential regulatory scenarios, whether they are related to the EU AI Act or the recent Executive Order issued by President Joe Biden.
While the founders do not have to know everything, they do need to have considered the consequences of any potential regulatory obstacles. Governments are voicing their opinions on what data can and cannot be fed into AI models, and copyright disputes are currently taking place. These cases are sure to come up again.
Recognizing cybersecurity issues
AI innovation is surpassing cybersecurity, much like regulation. Companies must be mindful of the situations in which insecure, advanced AI could expose their corporate data. Due to security flaws with third-party software providers, there have already been several significant hacks, which has forced companies to reconsider how carefully they select their suppliers. Startups have to consider the cybersecurity requirements and concerns of their firm.
Within the organization, Gen AI is creating new assault avenues and surface regions. Adversarial attacks, rapid injections, data poisoning, and jailbreaking model alignments are just a few of the issues that still need to be resolved before large-scale deployment is secure, dependable, and resilient. Defensive strategies will undoubtedly include AI-infused cybertools, but safeguarding AI is a developing area of cybersecurity.
When AI founders show initiative in addressing cybersecurity and regulatory issues, it is a positive sign.
Why data dictates the fate of startups
Data is the most important determinant of a startup’s ability to endure over time and survive the hype cycle. For startups to generate long-term value, they need to be in charge of their data destiny. What is your data strategy? is a more pertinent question to ask than “what is your gen AI strategy?” since a company’s model is only as strong as the quality of its data. A person’s ability to obtain high-quality data determines their success or failure. An organization’s ability to gather, prepare, and get value from data—as well as its ability to create a data flywheel—is a crucial component of its success.
The incapacity to gather and prepare the necessary datasets in a business is the main reason why most enterprise AI projects fail. Another issue is that many use cases in the industry will not initially have the benefit of internet-scale datasets. This gives artificially generated data the chance, at least sometimes, to force-multiply any data that organizations have access to.
For a number of years, this field has been intriguing, and it still offers potential for discoveries that could lead to a feedback loop where artificial intelligence models are improved by synthetic data. Notable instances of this are beginning to appear at the nexus of simulation tools, general artificial intelligence, and autonomous vehicle development. Similar strategies could be seen in foundation models that are more vertically oriented.
Where will the hype cycle around AI lead?
It seems obvious that advancements in Gen AI will keep coming in waves, and that software and APIs will keep maturing in short bursts. Claude 3, GPT-5, and Sora are just a few examples of the models that will continue to excite us as they show notable improvements in capacity. Like previous hype cycles, we have to face the fact that, although highly promising, emerging technology does not provide us with a whole picture, and we cannot draw hasty judgments on what the next AI wave means for every field.
To gain an understanding of the direction the industry is taking, I would contend that we should listen to academics, builders, and doers rather than venture capitalists, who are, quite simply, more adept at selecting businesses than long-term trend forecasters.
