Welcome to AI’s Trough of Disillusionment. In a little over a year, we have gone from Elon Musk calling for a general halt on AI development to Goldman now claiming that AI will become a major boondoggle. Frankly, most people who do not have fluidity with the tools available don’t yet understand what the technology is capable of, nor how to implement it into their workflows. Just because we have the technology doesn’t mean that people will automatically run to it or can immediately adopt it.

Most organizations have either haphazardly implemented it or thought carefully about where and what jobs AI is best suited for. In McKinsey’s Global Survey on AI, and nearly one-quarter of organizations are already experiencing negative consequences from generative AI inaccuracy. For instance, Air Canada faced issues with an AI chatbot that provided misleading information to customers and lost a legal case; a pharmacist review study found that nearly 75% of responses from ChatGPT for medication questions were found to be incomplete or wrong, and 56% of surveyed companies emphasized the risk of biases and hallucinations affecting the quality of AI outputs.

The issue stems from how to know what workflows can be created, how they can be created, and what their limitations are. Take for example, this publication. While endeavoring to write the opening section with my own words (ChatGPT writes like a precocious 12th grader that I find grating), I nonetheless figured out how to automate the tedious process of summarizing articles and then hyperlinking back to them for this newsletter. This wasn’t a painful process and didn’t take much skill to do. But, by doing so, I cut at least three hours of grunt work But just creating it and “hoping for the best” was not the optimal solution.

Here is what I did: I had ChatGPT rewrite each generated string so it was syntactically correct. Subsequently, I added the HTML marker tags to the descriptions. Finally, I pasted the responses below, which were then perfectly written and formated. If that isn’t significant value and time savings, I am not sure what else could be.

The difference between what I built and what other people have implemented is that:

• I knew what the end result was that I wanted.

• I could design a system to consistently generate the output I wanted connected to the original text.

• The task was primarily an information-parsing exercise focused on functions that LMMS does well in.

If you are curious, you can see and freely use the first push here. Personally, I have never thought that there was going to be AGI, but LLMs are truly revolutionary. Just remember that Uber took about two years to emerge after GPS was put on the iPhone. It will take time for the real use cases and developed applications to emerge for this technology.

As for other news, I am working on a presentation on how to evaluate startups and startup opportunities and have a couple of other irons in the fire.

News I Have Been Reading

• Sequoia Capital and Andreessen Horowitz have conflicting views on the demand for GPUs. Andreessen is stockpiling GPUs, believing the shortage will persist. Sequoia thinks the peak has passed and increased production will burn cash. Andreessen aggressively invests in AI startups, leading deals worth $1.3 billion, including OpenAI competitors. Sequoia focused on smaller, early-stage investments totaling $400 million. Read in The Information 

• Is open source winning the AI race? Since early 2022, $7.5 billion has been invested in open-source developers, including Hugging Face. Read in The Information

• Are we at the end of the model performance race? The transition in AI strategy from focusing on model performance to product development. Read in Generative AI  

• for those living under a rock, Former President Donald J. Trump was wounded in an assassination attempt during a rally in Butler, Pa. A gunman opened fire, injuring Trump and leading to the death of a spectator and critical injuries to two others. Read in The New York Times

• Goldman Sachs Research explores whether the massive investments (~$1tn) in generative AI technologies by tech giants will yield substantial benefits or returns. The article examines potential outcomes and implications for economies, companies, and markets. Read in Goldman Sachs Research 

• Houston, We finally have M&A! Google parent Alphabet is in advanced discussions to acquire cybersecurity startup Wiz for roughly $23 billion. If completed, it would be Alphabet's biggest acquisition ever. The deal is being considered amid antitrust scrutiny and could bolster Alphabet's cloud computing business. Wiz's investors include Insight Partners, Blackstone, Index Ventures, Cyberstarts, Sequoia, Greenoaks Capital Partners, Salesforce Ventures, and Lightspeed. Read in Seeking Alpha 

• How Neobank Synapse's bankruptcy left many fintech clients and their users stranded with approximately $160 million in frozen funds. Attempts to rescue and restructure the company failed, forcing a complete liquidation. The timeline of events reveals major setbacks and efforts to reconcile the situation, adversely impacting many fintech companies and their customers. Read in TechCrunch 

• Companies are searching for results that demonstrate generative AI's business value. While companies are pressured to invest in AI to stay competitive, the tangible business metrics proving its value remain elusive. CIOs are wary of adopting expensive technologies without clear returns on investment, leading to a dilemma of trust versus necessity. Read in TechCrunch

• While others flounder using AI, Amazon strengthens its AI capabilities by hiring leaders from Adept, a startup focused on automating enterprise workflows, while integrating some of Adept's technologies into Amazon's initiatives. Read in GeekWire 

Podcast Episodes

None this week; everyone is on vacation. I will have more next!

Raises

It is a slow week and will likely be so till after Labor Day. But for the worthwhile announcements, here they are:

• Hayden AI conducts research and development to accelerate growth, with support from investors like The Rise Fund, Drawdown Fund, and Autotech Ventures. It raised $90 million with The Rise Fund as the lead investor.

• Standard Bots, a pioneer in robotics automation, raised $63 million with the lead investor, General Catalyst.

• Buildots is a global construction technology leader, bringing the power of AI and computer vision to automate on-site progress tracking. The platform provides construction teams with accurate data and previously unavailable predictive performance metrics, empowering them to make data-driven decisions. Buildots' platform improves project visibility and efficiency, leading to reduced delays, enhanced error detection, and successful project delivery. Buildouts is helping increase the efficiencies of large-scale construction projects across North America, the UK, and the EMEA. It raised $15M with the lead investor, Intel Capital.

• HerculesAI, a HerculesAI has been working at automating professional services since 2017, focusing on the legal industry. The company has been building large language models for several years. It raised $26 million with the lead investor Streamlined Ventures.

• Enso gives small and medium-sized businesses (SMBs) access to preprogrammed AI agents that handle repetitive tasks; raised $6 million with the lead investor, NFX.

Final Thoughts

That’s it for this week. The podcast will soon have significant updates and exciting guests. Stay tuned! Please like and subscribe.