2023 AI Year in Review + 2024 Predictions

AI with ALLIE

The professional’s guide to quick AI bites for your personal life, work life, and beyond.

January 5th, 2023

Today, I’ll be sharing with you my 2023 AI Awards, AI Lowlights, and my 2024 Predictions 🔮 

This year was huge for AI, so I wanted to make sure you have direct access to these insights and stay ahead of all the fast-moving developments in the AI age.

So, without further ado, let’s get to it.

2023 was the year of AI.

And the evolution has been both fast and furious.

It started with ChatGPT, which attracted more users than the entire population of Vietnam in under two months. Since that launch, we’ve seen AI explode (there are over 40K AI companies), trend (like OpenAI CEO’s boomerang and Google Gemini chaos), and make its mark (see: Nvidia’s stock 📈).

🏆 My 2023 AI Awards 🏆

I went through all of my notes, emails, posts, and texts (and occasionally vivid nightmares) to recap all of the 2023 AI greats across 10 major categories. And it goes without saying that 2023 was momentous for AI. It was the hardest year yet to pick winners (the runners up and winners were neck-and-neck in some categories, like “productivity” and “dev tools”) and try to keep honorable mentions to a reasonable length.

Note: These are only my opinions and you may disagree, there will be some overlap in these categories (ChatGPT could be in almost all of these), and inclusion on this list is not an endorsement.

👎 My 2023 AI Not-So-Greats 👎

Not everything was AI rainbows and sunshine. (Unless of course, you’re a Midjourney addict, then literally everything can have AI rainbows and sunshine added.) In any field, setbacks allow for growth and increased learning—not just for the companies impacted, but for all those watching.

Every lowlight listed here is a lesson—about resilience, the courage to pivot, or even just what not to do. Sometimes, a company (or you yourself) needs to take a step back and recalibrate to keep moving forward with greater strategy.

🔮 And Lastly, My 2024 AI Predictions 🔮

Reminder: it was part of my literal job for years at Amazon to predict the future of AI. Also reminder of the famous philosophy quote:

I’m in the camp of predicting more gradual AI improvements rather than massive breakthroughs this year - which is an easier bet coming off of a year of crazy releases, rumors, and developments. With that, I wouldn’t wait for “the next best thing” to come around.

Assume new rollouts this year will not trump what’s out there today by 10x AND be both hard and expensive to adopt. Either the cost to switch will be low-ish, or it won’t be worth switching to.

In short: get your AI strategy together now.

Here are my predictions and explanations of each one. Take the time to read through this to have a more grounded 2024.

• LLMs optimized for structured data; analytics continues to be under-appreciated in AI (Code Interpreter in ChatGPT is life-changing for me); I’m expecting more LLMs for time series like TimesNet/TimesGPT (Microsoft already partnered with Nixtla!), general forecasting, or structured data predictions; I would not be surprised if there were more no-code releases like SageMaker Canvas to build your own model.

• Rise in system 1 + system 2 combo; my favorite book recommendation is “Thinking Fast and Slow” by Daniel Kahneman, and we’re going to see more balance of system 1 (fast! emotional! instinct!) and system 2 (slow, logical, analytical) AI systems. Here’s a great “what is AI” infographic from Aspen Institute on what that means.

• AI becomes more of an active contributor in decision-making and reasoning; the CICERO paper from Meta continues to be a favorite of mine (and the lead author now works at OpenAI), startups like Adept AI and Imbue Systems have raised hundreds of millions of dollars; more consistent and robust decisions, enterprise decision collaboration opportunities.

• Increased VC investment in LLM Ops; we saw the same thing happen with traditional ML—platforms for ML (like Amazon SageMaker, Weights and Biases, Vertex AI, and Scale AI) and then a rise in ML ops (like Fiddler AI and Tecton AI). The same thing will happen in LLMs like Amazon Bedrock, Llama Index, Last Mile AI, and Datology.

• Productionized agentive AI and new “teams” or “swarms” of AI agents; AI will improve at knowing and move into doing, more agentive AI that will take actions and execute multi-step tasks (ex: shop for and buy you a couch, code and deploy an app, talk to and integrate with another software), more exploration of agents as tools and swarms (like “mixture of experts” approach but in agentive form); early examples include Hyperwrite’s AI Assistant (a ChatGPT that actually takes actions for you) and Amazon Bedrock Agents.

• We will start to uncover what comes after transformers, after generative AI, and even after the AI age; have posted predictions before, refer to my previous long-form post on the “evolution age”; also will be keeping my eye on Liquid AI from the AI icon Daniela Rus and the MIT spinout’s liquid neural networks.

• Several AI startups will raise billions of dollars; obvious picks are AI heavyweights like Open AI (prediction: will pass $100B valuation), Anthropic (prediction: will pass $50B valuation), Inflection AI (the makers of Pi, already raised over $1B), Adept AI, Character AI, Hugging Face, Cohere AI; that said, overall VCs will have a higher bar for generative AI startup investments (i.e. strong competitive moat, clear path to monetization).

• We will see a fully AI-generated album, TV show (even a news station), or movie, and it will probably not be well received by the arts community; tension between the creativity community and tech will continue. Keep an eye on Suno AI for music, RunwayML and Pika for video.

• AI will start to become more integrated into our bodies; big hardware battles for customer attention with several big consumer hardware rollouts (e.g., Apple Vision Pro, Humane AI, Rewind AI, Tab); first true tests of whether smartphones are replaceable in the next 15 years, Apple Watch patent issues with Apple might slow things down for them; I remain bullish on XR and AR and bearish on VR.

• Early warning signs of AI and elections and likely more deepfakes; dozens of countries constituting half of the world’s population will have an election in 2024 (including India, USA, Mexico, and South Africa); expect more sophisticated deepfakes at a larger scale than we've seen to date. AI will be used to create disinformation and, often ignored, AI will be blamed for information that is true but unwanted.

• Big AI launches from “traditional” tech and non-tech large enterprise; everyone knows that Microsoft, Google, Amazon, and Meta are investing heavily in AI, and we’ve seen ISVs like Salesforce, ZenDesk, and ServiceNow launch their own AI-enabled tools alongside more experimental launches like Coca Cola’s AI-generated flavor in September and Carvana’s AI ads. This year, expect more meaningful AI releases from other tech players (like AMD, Intel, Apple, Samsung, Dropbox) as well as large traditional enterprises like Home Depot, United Health, JPMC, Visa, Mastercard, Walmart. These companies have likely been working behind-the-scenes on AI, wanting to launch internal-first for risk mitigation.

• Compute continues to be like gold; Nvidia is throwing big money into startups and creating partnerships (like Imbue Systems), Amazon threw “up to” $4B into Anthropic for the same, and Alphabet/Google invested “up to” $2B in Anthropic as well. My best guess (based on working at two cloud providers and reading the public press releases) is that these deals require these startups to use the company’s chips/compute for specific workloads and act as a flagship marketing reference for their technology. AMD/Rain/MatX/Cerebras will likely follow the same playbook, Microsoft will start building out their chip group and follow suit, Meta already made 2 chips and will do the same in the open source space.

• Newer ML cloud threats from both hardware makers and startups; there was a big push from startups like Foundry, CoreWeave, Banana.dev, Tensorwave to capture the “AI in the cloud” market in 2023 with purpose-built clouds for ML workloads, mega AI startups like will move down the stack (like Inflection AI announcing their supercomputer in July 2023), and hardware makers like Nvidia will also continue to move up the stack (Nvidia has been trying to move into more cloud action for years).

• More decentralized AI and more open source AI options; closed models will continue to outperform open source models in 2024 (there is large disagreement on this from AI thought leaders, but it’s important to note that the OSS community is saying they will be equal and the closed source is saying it will continue to outperform OSS; most “unbiased third parties” like myself say closed will outperform).

• Smaller AI models make a huge splash; AI models that can wholly live on phones (like Google Gemini Nano and rumored AppleGPT) and OSS small LLMs from Mistral AI and Meta will have their moment.

• More hyperpersonalization, more robust prediction power, and more proactive systems; the eventual future is fully personalized blog posts, movies, books, and every type of content. That begins now (as an example, look to companies like OpenAds which supplies personalized advertising within your AI assistant chat interface, tailored to the conversation you’re having). And something I’ve been shouting for years: AI will help with interventions, both flagging them and taking action on them.

• Bigger AI scams, so prepare your company and family; you can create an AI clone of someone with just a few videos they’ve posted online. Scammers will create new tricks like cloning someone and autodial their family at scale (reminder: create a password with your friends/family to confirm your identity - never type it out, and once it’s used once, create a new one).

• First tunable releases in both value-oriented and goal-oriented AI (I give it a value system, it tunes a model, or I give it a goal, and it tunes the outputs towards that goal); not enough has been said about constitutional AI from Anthropic, and I think it’s the future.

• Digital entrepreneurship skyrockets; the ability for anyone to build a company, product, solution continues to get easier and the effort required to get it done drops dramatically. It’s not crazy to think of billion-dollar companies being built by teams of under 10 people (Midjourney supposedly hit $200M ARR with 11 employees and $0 investment, but likely struck a rev share deal with Discord to get there); more entrepreneurial compression and it has never been more important to find your squad of 6.

• More politically-driven AI and battles over model “wokeness”. Sam Altman’s request for OpenAI feedback had ‘remove wokeness’ as a top ask and Elon’s/Grok’s marketing messages for X’s AI push are heavy on it. We will see more brand pages will talk about “wokeness” versus “freedom”, with loud political undertones. Whether it’s the product itself or a tunable option in the product, we will see more “red vs blue” options like Neeva originally teased out.

• Companion AI will become its own app category with players like Character AI, Inflection AI, Replika, Digi AI; romantic relationships with AI will still be low/early adoption, but by EOY, it will be slightly less taboo (think: therapy for boomers vs millennials).

• More AI-generated influencers like Aitana Lopez, Emily Pellegrini, and Sarah Jordan, largely sexualized content, created largely by men, largely for men.

• More real people create AI twins of themselves, more easily (like Sika Moon) to “clone” themselves and create more content, in a personalized way, faster, and for global audiences. And the company that wins in this space (Meta? Character AI? Someone else?) will be like Facebook in 2006.

• A few celebrities will make splashes in artificial intelligence apps, perhaps by launching their own mobile apps or web apps, likely targeted in the content creation/influencer, media and entertainment, retail/ecommerce space, with at least one big win.

• More kid-specific apps and toys, some with a creepy factor of >0; some will be AI education tools like Kyron Learning (customized tutoring, launched by the founder of Gmail) and QPad AI (teaching kids about AI), and some will be more for entertainment like Curio (which is like a Furby + ChatGPT); parents will have to make difficult decisions over what AI tools and products their children can use and what their family AI rules are (just like how parents pick when their child can get a phone or social media) - this is something I will likely make some posts on this year because I think too many parents will jump into it, not realizing that their kid is being tracked.

• More real-time AI as costs drop; real-time, even in non-AI land, is usually one of the later releases because it requires new processes, architectures, and cost structures (the order of releases is usually: single modal, single modal and high performance, multi-modal, multi-modal and high performance). I imagine as costs drop and as OSS AI model performance increases, we will be able to see more real-time AI like Krea AI’s real-time image transfer.

• People will keep thinking we’ve achieved AGI when we haven’t; companies need to get more clear on a definition and benchmarks (also, in general, AI benchmarking and comparisons are still muddy), Twitter/X will continue to be filled with crypto-fiends-turned-AI-fiends saying we’ve achieved AGI and arguing in the comments, note that this mostly matters if there are consequences to the binary question “have we achieved AGI” like regulation, and not general impact (like “can this alter the banking system”).

• Multimodal AI means language to…everything, and it’ll grow; natural language is just an interface, and chat isn’t everything. I also think we will see much more voice-enabled tech (I talk-prompt 20-40 times a day); beyond new inputs, expect a rise in outputs that happen away from the screen (like typing to control a robotic arm); eventually, code systems will talk to each other and auto-connect and auto-improve. Yes, code-to-code or AI-to-AI or system-to-system will be huge (with early signs in 2024). And if you wanna hear something really wacky, I think we will be able to talk to our pets in under 10 years.

• Way WAY more AI jobs, and not just ML engineers; in the US alone there are 56K open jobs with the phrase “artificial intelligence” (on a LinkedIn search). I’m predicting AI roles at the leadership level to ramp up, like the search for AI policy leads, Director of AI (currently open at LA Clippers, if you’re interested), and Chief AI Officer/Head of AI (several prominent Fortune 500 companies, private equity firms, and financial institutions are on the hunt), as well as more generalized/newer positions like AI operations, AI product managers, AI project managers. If you’re hiring in AI, let me know so I can share it with my audience.

• And many, many more predictions, like: more domain-specific AI and SaaS since that’s harder for big tech to compete on, more AI builders, more people working in AI, more people shirking traditional job roles either voluntarily or involuntarily (in order to start their own company, work freelance, side hustles, social media influencers, non-traditional roles, layoffs, remote work guarantees, gig work - I have a really hard time imagining traditional corporate roles increase in popularity), people voicing not wanting to join companies that ban AI, more discussions on the “cost” (not just money!) and overall sustainability (not just environmental!) of AI, more no-code platforms and adoption…

2024 is going to be just as action-packed. I’m predicting everything from new AI agent ‘teams’ to election interference to more AI decacorns to compute battles to step changes in worker productivity.

Most business leaders have “woken up” to the AI revolution, but unfortunately, very few actually know what to do and how to take advantage of it.

I hope your new year is off to a great start and your 2024 will be filled with only the best AI has to offer. And if you want to get help or work together this year, email me and my team at [email protected] or just fill this out.

New ‘AI for Business Leaders’ Cohort Starts March 2024

If all this makes you feel a bit behind in the AI game, don't sweat it. I've got you covered with my upcoming ‘AI for Business Leaders’ course where you’ll learn how to apply AI into your business.

Here’s what you can expect:

• Contextualizing AI into your business and the environment today

• Finding your first (or second or third…) AI use case

• Methods to evaluate and prioritize use cases

• My favorite AI examples for enterprise and personal productivity

• Real-world tactics of ML execution, including business trade-offs

• Change management for AI

• Responsible AI deployment and risk mitigation

• How building responsible AI can enhance customer trust and value

• My favorite AI tools and resources for business leaders

The waitlist is now open and my next cohort is set to launch in March 2024, so make sure to grab your spot today.

Feedback is a Gift

I would love to know what you thought of this newsletter and any feedback you have for me. Do you have a favorite part? Wish I would change something? Felt confused? Please reply and share your thoughts or just take the poll below so I can continue to improve and deliver value for you all.

Stay curious, stay informed,

Allie