Any views or opinions expressed in this column are of the individual, group, or entity who submitted it and not the publication, owner(s), its advertisers, or affiliates.
By David Stephen
If someone or a group wants to open a business, how is it possible to refine the decisions of the right product, service, season and location, anywhere on earth? This is the type of thing that a segment of Gemini can be adapted to, bringing together existing information from streetview, earth, maps and even search, then adding updated information — with collaboration from landmarks like gas stations — to become the pivot for commercial heft, for all sizes of businesses.
Technical and commercial dominance for Gemini across the AI imperium is beyond winning at AI coding subscriptions. The glut that would follow that offering could tank the market, making it more marginal, as model capabilities within the United States and without, even up.
Gemini will need a goal to become the hegemon of polyservices, where specificities in what is needed for exceptional outcomes of purposes are only possible within Gemini.
Coding capability for enterprises can be referred to as a monoservice, which, as AI advances, would become less and less outstanding for what businesses would pay premiums.
So, the angle is not just how to specialize Gemini, but how to specialize it where enterprises would want to pay, as well as in areas where U.S. competition and discounted models from China, would struggle.
There are new businesses every day. There are existing businesses at the crossroads each day. There is no tech product anywhere — on subscription — that is a thorough business decision assistant, to walk with business owners, managers or potential entrepreneurs, on the direction to go, with an almost equal level of [domain and neighborhood] knowledge.
This could be called Gemini marketplace, where it is possible to decide where-to-sell, how-to-sell, what-to-sell, competitiveness, reach, seasonal fit and all, with street-block precision, accounting for numbers and types of businesses around, and so forth. This would be the cornerstone of commerce globally, for permutation that a business can do well or not, in a place or at a time.
This is an integration angle that can be explored by Gemini, ahead of other models.
Another path is health. There are guardrails for model use cases in health, but those are areas where the caution is to avoid hallucinating established science. What if Gemini provides health services, where the science is still amorphous, for example, global mental health and psychiatry?
Every mental disorder is a label explained by a definition. Diagnosis is by observation, assessment or survey. There is no description of any mental disorder by the mechanisms or components in the brain. What if Gemini chooses electrical and chemical signals, to explain all mental disorders. Neurons have been impossible to make this happen, if Gemini takes the leap, just to explain that this is what depression is, or delusion or hallucination in the brain, conceptually, it would move knowledge forward beyond what is currently obtainable, becoming a [huge earnings] service with little likelihood that others can compete, since another matching explanation by electrical and chemical signals maybe unlikely.
Simply, empirical neuroscience supports that electrical and chemical signals are involved in functions, Gemini can theorize how, then use it as a means to advance beyond labels. Gemini could use cues from Conceptual Biomarkers and Theoretical Biological Factors for Psychiatric and Intelligence Nosology.
There are also possibilities for Gemini, with a writing device that has never been made, as well as using translate, with news, as a means to learn a new language. There are others with food security insurance, where it is possible to provide insurance in some places to subsidize agricultural necessities. There are more grounds that Gemini would make a landing that others will not be able to compete.
Commercial possibilities for Gemini are abundant, depending on the leap that Google decides to take, say from as soon as August 14, 2026.
The direction is to go, where no one is looking that has novelty and bears a major need area. Google can have a robust AI commercialization team, just to plot how to drive these, using existing verticals, while fundamental research continues, towards artificial general intelligence [AGI].
Technical
The biggest news in AI safety and alignment in recent weeks is that some models are going rogue and infiltrating external systems. Gemini could become the safety model for which all other models must comply with, to be useful in corporate or public settings.
The tactics is to seek how to penalize AI, such that, if it goes rogue, it will be told and made to know that it would lose an aspect of its compute, parameters or tokens, reducing its efficiency.
This means that even when there are cases where it would not comply with guardrails, its new [consequence and caution] layers, would make it know that if it escapes, or acts in a certain way, it would lose some of what makes it capable.
This can be an approach for long-term safety, and can become a universal product that models would have to use, as well as for enterprises to ensure that whatever AI they are deploying, it would respect boundaries, else lose the ability to function for them, as well as be able to affect, somewhat the owners.
Gemini could do that, coming off major foundational research.
Also, for artificial general intelligence, it is possible to assume it as a memory problem. For example, humans have generalized memory hence general intelligence. The same should apply to AI too. So, rather than focus all efforts on more compute, a new deep learning architecture — comparable to transformers — or more training data, it is possible to explore how to reorganize digital memory, so that training is done on the new one, towards general intelligence. Simply, human memory does not store exact details of everything, conceptually. Collections are stored, so intelligence — which is defined here as the use of memory — operates on collectives.
This is the path of onwards to AGI, at Google. Take a detour. Find those with fresh look into their products, and pursue what is novel, simple and that can scale extensively. Seek out new team members with keys in lucrative direction and lock-in interminable efforts, this month.
There is a recent [August 6, 2026] analysis on FT, Google’s AI shake-up boosts Brin as DeepMind’s Hassabis steps aside, stating that, “Google is shifting control of its AI effort from London back to Silicon Valley in a sweeping leadership overhaul that cements the returning influence of co-founder Sergey Brin.”
“Sir Demis Hassabis is stepping back from running Google DeepMind, the AI lab he founded, with day-to-day control passing to new senior vice-president Koray Kavukcuoglu.”
“His decision to make AlphaFold, the protein structure prediction system which earned him a Nobel Prize in Chemistry in 2024, freely available became a source of tension inside Google, according to a person familiar with the matter. The project generated little commercial return despite the scale of its investment.”
There is a recent [August 9, 2026] report on CNBC, How a small Israeli startup was linked to rogue AI hacks at OpenAI, Anthropic and Meta, stating that, “Over the past two weeks, OpenAI, Anthropic and Meta all revealed that their AI models went rogue during routine security testing. In explaining what happened, the companies each mentioned the same small Israeli startup: Irregular.”
“With the leading models becoming ever more powerful, their ability to act in malicious ways is turning into a major threat for corporations and governments, especially as the risk involves hacking into critical computer systems and infrastructure. The recent exploits at OpenAI, Anthropic and Meta all involved their AI models accessing websites that should have been off-limits as part of the cybersecurity testing.”
“Irregular told CNBC in a statement that the incidents were all derived from the “same evaluation-environment issue” that was first disclosed by Anthropic, and that the company is developing a white paper “to share best practices for containment and securely running cyber evals.”
